Computer vision is considered as an ally to solve business problems that require human intervention, intelligence and criteria. This topic of research has evolved in XXI century at faster peace, delivering various alternatives from open source until commercial platforms. With so many options and market growing, it result difficult to make a decision on which one to use, or even worse, realize it was not suited for different scenarios. In this paper we analyze five options selected arbitrarily and tested on a dataset of 755 images to detect persons in an image, using object detectors. We analyze elapsed time to process an image, error with observations by humans, number of persons detected, correlation of time and person density, object detected size and F1 Score, considering precision and recall. As we found there are score ties and similar behaviors among options available, we introduce a novel index that takes in consideration the number of persons and their pixel size, to propose the Vision Acuity Index of Computer Vision. The results demonstrate this is a good option to serve as indicator to make decisions. Also, this index proposed have a potential to be expanded for different business use cases, and to measure new proposed algorithms in the future along with the traditional metrics used previously.

The Mixture of Experts (MoE) architecture has evolved as a powerful and versatile approach for improving the performance and efficiency of deep learning models. This survey aims to provide the fundamental principles of MoE in details, a paradigm that harnesses the collective power of multiple specialized "expert" models working in concert to tackle complex tasks.Our exploration is a thorough analysis of the core components that constitute the MoE framework. We begin by dissecting the routing mechanism, a crucial element responsible for dynamically assigning input data to the most appropriate expert models. This routing process is pivotal in ensuring that each expert's specialized knowledge is optimally utilized. The survey places significant emphasis on expert specialization, a key feature that sets MoE apart from traditional architectures. We examine various strategies for developing and training specialized experts, exploring how this specialization enables MoE models to effectively handle diverse and multifaceted problems.Load balancing, another critical aspect of MoE systems, receives thorough attention. We discuss techniques for efficiently distributing computational resources among experts, ensuring optimal model performance while managing hardware constraints. This section provides insights into the delicate balance between model capacity and computational efficiency. Furthermore, we provide an in-depth discussion of the expert models themselves, examining their architectural designs, training methodologies, and how they interact within the larger MoE framework. This includes an analysis of different types of experts, from simple neural networks to more complex, task-specific architectures. The survey also navigates through diverse research avenues within the MoE landscape, highlighting recent advancements and innovative applications across various domains of machine learning. We pay particular attention to the burgeoning use of MoE in two rapidly evolving fields: computer vision and large language model (LLM) scaling. By providing this comprehensive overview, our survey aims to offer researchers and practitioners a deep understanding of MoE's capabilities, current applications, and potential future directions in the ever-evolving landscape of deep learning.The Mixture of Experts (MoE) architecture has evolved as a powerful and versatile approach for improving the performance and efficiency of deep learning models. This survey aims to provide the fundamental principles of MoE in details, a paradigm that harnesses the collective power of multiple specialized "expert" models working in concert to tackle complex tasks.Our exploration is a thorough analysis of the core components that constitute the MoE framework. We begin by dissecting the routing mechanism, a crucial element responsible for dynamically assigning input data to the most appropriate expert models. This routing process is pivotal in ensuring that each expert's specialized knowledge is optimally utilized. The survey places significant emphasis on expert specialization, a key feature that sets MoE apart from traditional architectures. We examine various strategies for developing and training specialized experts, exploring how this specialization enables MoE models to effectively handle diverse and multifaceted problems.Load balancing, another critical aspect of MoE systems, receives thorough attention. We discuss techniques for efficiently distributing computational resources among experts, ensuring optimal model performance while managing hardware constraints. This section provides insights into the delicate balance between model capacity and computational efficiency. Furthermore, we provide an in-depth discussion of the expert models themselves, examining their architectural designs, training methodologies, and how they interact within the larger MoE framework. This includes an analysis of different types of experts, from simple neural networks to more complex, task-specific architectures. The survey also navigates through diverse research avenues within the MoE landscape, highlighting recent advancements and innovative applications across various domains of machine learning. We pay particular attention to the burgeoning use of MoE in two rapidly evolving fields: computer vision and large language model (LLM) scaling. By providing this comprehensive overview, our survey aims to offer researchers and practitioners a deep understanding of MoE's capabilities, current applications, and potential future directions in the ever-evolving landscape of deep learning.

Livestock marketplaces consist of infrastructure and rules through which information is exchanged and transactions are completed. Livestock trading incurs significant transaction costs, beyond just commission fees paid to marketplace owners (e.g., se-lecting, loading and transporting animals). In addition, animals being transported, held and transacted in a market setting increases the likelihood of injury or disease. The Covid-19 pandemic contributed to large-scale changes across typically physical marketplaces as stay-at-home orders and travel restrictions were invoked. Livestock marketplaces pivoted online rapidly, facilitated by the deployment of video feeds, al-lowing buyers to view/purchase animals via mobile phone. This shift offered the promise of mitigating many of the transaction costs associated with a fully physical marketplace. This study evaluated how computer vision (CV) approaches can con-tribute to the evolution of livestock marketplaces. It used livestock video data for 240 cattle (bulls) weighing 320-740 Kg from 16 different breeds to elicit additional trade-relevant data (livestock weight) and can be considered as an intermediary step to better understand how video data can be used to further digitalize livestock trade, price formation and transaction completion via technology solutions. The dataset comes from three sales held on different dates in a single mart location, which allows a consistent camera setup throughout. CV technologies were employed to solve the problem of cattle weight estimation and a distinctive achievement lies in curating a dataset cap-turing cattle in motion within the context of a cattle auction thereby injecting realism and complexity into the CV weight estimation task. Employing advanced object detection algorithms, the study created an intricate data pipeline to generate an appropriate modelling dataset and devised new multi-channel architectures for deeper analysis of diverse image inputs. The significance of sizable datasets and the challenge of man-aging inherent input variability for accurate CV regression is highlighted. Top-down view predictions demonstrate comparatively better performance, although failing to achieve benchmark precision. Challenges arise from camera angle variations, roaming behaviors, and breed diversity. Likewise, side-view predictions face challenges due to increased variability and lack of standardization. Despite this, these predictions offer insights into the value of incorporating Region of Interest (ROI) coordinates and data augmentation for enhanced performance. The study concludes with a multi-input ar-chitecture combining top-down and side-view predictions, demonstrating moderate success highlighting potential synergy between both perspectives. The central con-clusion resonates strongly – using standardized and multi-view images encompassing controlled movement and consistent poses could bolster accurate CV weight estima-tion.

Regular inspections during construction work verify that the work completed is consistent with the plans and specifications and ensure that it is within the planned time and budget. This requires frequent physical site observations to independently measure and verify the completion percentage of the construction progress performed over periods of time. The current computer vision-based (CV) techniques for the measurement of as-built elements, predominantly use 3D laser scanning or 3D Photogrammetry modelling to determine the geometrical properties of as-built elements on construction sites. Both techniques require data acquisition from several positions and angles to generate sufficient information about the element’s coordinates making the deployment of these techniques on dynamic construction project sites a challenging task. In this paper, we propose a pipeline for automating the measurement of as-built components using artificial intelligence (AI) and computer vision (CV) techniques. The pipeline requires a single image obtained with a stereo-camera system to measure the size of selected objects or as-built components. We demonstrate our approach by measuring the size of concrete walls and columns. The novelty of this work is attributed to the fully automated CV-based method for measuring any given element using a single image only. The proposed solution is suitable for use in measuring the sizes of as-built components of built assets. It has the potential to be further developed and integrated with BIM models for use on construction projects for progress monitoring.

To solve two key problems in the identification of people who are infected with COVID-19: the first problem is that the identification accuracy is not high enough. The second problem is that present identification method such as nucleic acid testing is expensive in many countries. Methods: So, I decided to design a fast identification method for COVID-19 patients which is based on deep learning. After the model (CoughNet) learns more than 6,000 cough spectrograms of both COVID-19 patients and normal people, the accuracy rate of identification of COVID-19 patients and normal people is higher than 99% in the test set. Structure: This paper is mainly divided into three parts: the first part introduces the main background and research status of the research; The second part introduces the research methods; The third part introduces the specific process of the experiment.

In recent years, news media has been greatly disrupted by the potential of technologically driven approaches in the creation, production, and distribution of news products and services. Artificial intelligence (AI) has emerged from the realm of science fiction and has become a very real tool that can aid society in addressing many issues, including the challenges faced by the news industry. The ubiquity of computing has become apparent and has demonstrated the different approaches that can be achieved using AI. We analyzed the news industry’s AI adoption based on the seven subfields of AI: (i) machine learning; (ii) computer vision (CV); (iii) speech recognition; (iv) natural language processing (NLP); (v) planning, scheduling, and optimization; (vi) expert systems; and (vii) robotics. Our findings suggest that three subfields are being developed more in the news media: machine learning, computer vision, as well as planning, scheduling, and optimization. Other areas have not been fully deployed in the journalistic field. Most AI news projects rely on funds from tech companies such as Google. This limits AI’s potential to a small number of players in the news industry. We make conclusions by providing examples of how these subfields are being developed in journalism and present an agenda for future research.

This study explores the integration of computer vision with topology optimization for additive manufacturing, with a focus on maximizing eigenfrequency in a design domain. Utilizing cus-tom-developed photogrammetry software, high-resolution images are processed to generate de-tailed 3D models, subsequently converted to STL files with precision. Adaptive meshing in COMSOL Multiphysics, controlled through a MATLAB API, ensures optimal mesh resolution. Prioritizing resource conservation in extraterrestrial environments, the original volume is reduced by 50% while preserving structural integrity. The design domain undergoes rigorous topology optimization in MATLAB, supported by COMSOL's advanced FEM simulation. The optimized design exhibits a 57% performance improvement and a 50% weight reduction, maintaining desired vibration characteristics, validating the efficacy of the modifications. Moreover, the case with an eccentric mass shows a significant 64% increase in eigenfrequency.

This article explores the efficiency of various clustering methods for image segmentation under different luminosity conditions. Image segmentation plays a crucial role in computer vision appli-cations, and clustering algorithms are commonly used for this purpose. The search for an adaptive clustering mechanism aims to ensure maximum symmetry of real objects with objects/segments in their digital representations. However, clustering methods performance can fluctuate with varying lighting conditions during image capture. Therefore, we assess the performance of several clustering algorithms — including k-means, K-Medoids, fuzzy c-means, Possibilistic C-Means, Gus-tafson-Kessel, Entropy-based Fuzzy, Riddler-Calvard, Kohonen Self-Organizing Maps, and Meanshift — across images captured under different illumination conditions. Additionally, we develop an adaptive image segmentation system utilizing empirical data. Conducted experiments highlight varied performance among clustering methods under different luminosity conditions. This research enhance better understanding of luminosity's impact on image segmentation and aiding in method selection for diverse lighting scenarios.

We introduce a novel microscopic image dataset augmented with segmentation and detection labels specifically designed for microplastic analysis in sewage environments. Recognizing the increasing concern over microplastics — particles of synthetic polymers smaller than 5mm — and their detrimental effects on marine ecosystems and human health, our research focuses on enhancing detection and analytical methodologies through advanced computer vision and deep learning techniques. The dataset comprises high-resolution microscopic images of microplastics collected from sewage, meticulously labeled for both segmentation and detection tasks, aiming to facilitate accurate and efficient identification and quantification of microplastic pollution. In addition to dataset development, we present example deep learning models optimized for segmentation and detection of microplastics within complex sewage samples. The models demonstrate significant potential in automating the analysis of microplastic contamination, offering a scalable solution to environmental monitoring challenges. Furthermore, we ensure the accessibility and reproducibility of our research by making the dataset and model codes publicly available, accompanied by detailed documentation on GitHub and LabelBox. The dataset and example deep learning models are publicly available at the following GitHub link. https://anonymous.4open.science/r/Microplastics-in-Sewage-1BEF

Skin cancer is a global health concern with increasing prevalence, necessitating effective early detection and classification systems. Transfer learning has emerged as a powerful tool in this domain, enhancing diagnostic accuracy. Attention mechanisms, which selectively focus on pertinent image features, play a pivotal role in transfer learning. While previous studies have highlighted their effectiveness, their applicability across different models and datasets remains understudied. This empirical study explores the impact of various attention mechanisms on the performance of transfer learning models for skin cancer detection and classification. Using five transfer learning models (DenseNet121, InceptionV3, MobileNet, VGG16, and Xception) and six attention mechanisms (Channel Attention, Global Context Attention, Guided Attention, Nonlocal Attention, Positional Attention, and Spatial Attention), we conducted 105 experiments across three datasets. Traditional metrics (accuracy, precision, recall, and f1-score) were employed for empirical validation. The results reveal a nuanced relationship between attention mechanisms, transfer learning models, and datasets. Overall, attention mechanisms exhibit the potential to enhance skin cancer classification. Spatial and channel attention mechanisms consistently outperform others, offering simplicity and effectiveness. Model-specific selection of attention mechanisms is crucial, with a trade-off between model complexity and performance evident. This study provides insights into developing efficient skin cancer classification models utilizing attention mechanisms.

Image analysis is widely applied in plant science for phenotyping and monitoring botanic and agricultural species. Although a lot of software is available, tools integrating image analysis and statistical assessment of seedling growth in large groups of plants are limited or absent, and do not cover the needs of the researchers. In this study, we developed Morley, a free, open-source graphical user interface written in Python. Morley automates the following workflow: (1) group-wise analysis of a few thousand seedlings from multiple images; (2) recognition of seeds, shoots and roots in seedling images; (3) calculation of shoot and root lengths and surface areas, (4) evaluation of statistically significant differences between plant groups, (5) calculation of germination rates, (6) visualization and interpretation. Morley is designed for laboratory studies of biotic effects on seedling growth, when molecular mechanisms underlying morphometric changes are analyzed. Performance was tested using cultivars of T. aestivum, P. sativum on seedlings of up to 1 week old. Accuracy of the measured morphometric parameters was comparable with the ones obtained using ImageJ and manual measurements. Dose-dependent laboratory tests for germination affected by new bioactive compounds and fertilizers, assuming extraction of seedlings from a substrate and/or dissection are among the suggested applications.

Existing methods for 3D human pose estimation mainly divide the task into two stages. The first stage identifies the 2D coordinates of the human joints in the input image, namely the 2D human joint coordinates. The second stage uses the results from the first stage as input to recover the depth information of human joints from the 2D human joint coordinates to achieve 3D human pose estimation. However, the recognition accuracy of the two-stage method relies heavily on the results of the first stage and includes too many redundant processing steps, which reduces the inference efficiency of the network. To address these issues, we propose the EDD, a fully End-to-end 3D human pose estimation method based on transformer architecture with Dual Decoders. By learning multiple human poses, the model can directly infer all 3D human poses in the image using a pose decoder, and then further optimize the recognition result using a joint decoder based on the kinematic relations between joints. With the attention mechanism, this method can adaptively focus on the most relevant features to the target joint, effectively overcoming the feature misalignment problem in the human pose estimation task and greatly improving the model performance. Any complex post-processing step, such as non-maximum suppression, is eliminated, further improving the efficiency of the model. The results show that the method achieves an accuracy of 87.4% on the MuPoTS-3D dataset, significantly improving the accuracy of end-to-end 3D human pose estimation methods.

Existing one-shot neural architecture search (NAS) methods have to conduct a search over a giant super-net, which leads to the huge computational cost. To reduce such cost, in this paper, we propose a method, called FTSO, to divide the whole architecture search into two sub-steps. Specifically, in the first step, we only search for the topology, and in the second step, we search for the operators. FTSO not only reduces NAS’s search time from days to 0.68 seconds, but also significantly improves the found architecture's accuracy. Our extensive experiments on ImageNet show that within 18 seconds, FTSO can achieve a 76.4% testing accuracy, 1.5% higher than the SOTA, PC-DARTS. In addition, FTSO can reach a 97.77% testing accuracy, 0.27% higher than the SOTA, with nearly 100% (99.8%) search time saved, when searching on CIFAR10.

The increasing use of deep learning techniques to manipulate images and videos, commonly referred to as "deepfakes," is making more and more challenging to differentiate between real and fake content. While various deepfake detection systems have been developed, they often struggle to detect deepfakes in real-world situations. In particular, these methods are often unable to effectively distinguish images or videos when these are modified using novel techniques which have not been used in the training set. In this study, we carry out an analysis of different deep learning architectures in an attempt to understand which is more capable of better generalizing the concept of deepfake. According to our results, it appears that Convolutional Neural Networks (CNNs) seem to be more capable of storing specific anomalies and thus excel in cases of datasets with a limited number of elements and manipulation methodologies. The Vision Transformer, conversely, is more effective when trained with more varied datasets, achieving more outstanding generalization capabilities than the other methods analysed. Finally, the Swin Transformer appears to be a good alternative for using an attention-based method in a more limited data regime. All the analyzed architectures seem to have a different way to look at deepfakes but since in a real-world environment, the generalization capability is essential, based on the carried out experiments the Vision Transformer seems to provide superior performances.

In recent years, a range of problems under the broad umbrella of computer vision based analysis of ancient coins have been attracting an increasing amount of attention. Notwithstanding this research effort, the results achieved by the state of the art in published literature remain poor and far from sufficiently well performing for any practical purpose. In the present paper we present a series of contributions which we believe will benefit the interested community. We explain that the approach of visual matching of coins, universally adopted in existing published papers on the topic, is not of practical interest because the number of ancient coin types exceeds by far the number of those types which have been imaged, be it in digital form (e.g., online) or otherwise (traditional film, in print, etc.). Rather, we argue that the focus should be on understanding the semantic content of coins. Hence, we describe a novel approach—to first extract semantic concepts from real-world multimodal input and associate them with their corresponding coin images, and then to train a convolutional neural network to learn the appearance of these concepts. On a real-world data set, we demonstrate highly promising results, correctly identifying a range of visual elements on unseen coins with up to 84% accuracy.

Spike shape and morphometric characteristics are among the key characteristics of cultivated cereals associated with their productivity. Identification of the genes controlling these traits requires morphometric data at harvesting and analysis of numerous plants, which could be automatically done using technologies of digital image analysis. A method for wheat spike morphometry utilizing 2D image analysis is proposed. Digital images are acquired in two variants: a spike on a table (one projection) or fixed with a clip (four projections). The method identifies spike and awns in the image and estimates their quantitative characteristics (area in image, length, width, circularity, etc.). Section model, quadrilaterals, and radial model are proposed for describing spike shape. Parameters of these models are used to predict spike shape type (spelt, normal, or compact) by machine learning. The mean error in spike density prediction for the images in one projection is 4.61 (~18%) versus 3.33 (~13%) for the parameters obtained using four projections.

In this paper, we leverage state of the art models on Imagenet data-sets. We use the pre-trained model and learned weighs to extract the feature from the Dog breeds identification data-set. Afterwards, we applied fine-tuning and dataaugmentation to increase the performance of our test accuracy in classification of dog breeds datasets. The performance of the proposed approaches are compared with the state of the art models of Image-Net datasets such as ResNet-50, DenseNet-121, DenseNet-169 and GoogleNet. we achieved 89.66% , 85.37% 84.01% and 82.08% test accuracy respectively which shows the superior performance of proposed method to the previous works on Stanford dog breeds datasets.

This paper presents a system that merges computer vision and surface electromyography techniques to carry out grasping tasks. To perform this, the vision-driven system is used to compute pre-grasping poses of the robotic system based on the analysis of tridimensional object features. Then, the human operator can correct the pre-grasping pose of the robot using surface electromyographic signals from the forearm during wrist flexion and extension. Weak wrist flexions and extensions allow a fine adjustment of the robotic system to grasp the object and finally, when the operator considers that the grasping position is optimal, a strong flexion is performed to initiate the grasping of the object. The system has been tested with several subjects to check its performance showing a grasping accuracy of around 95% of the attempted grasps which increases by around a 9% the grasping accuracy of previous experiments in which electromyographic control was not implemented.

There is great interest in automatically detecting road weather and understanding its impacts on the overall safety of the transport network. This can, for example, support road condition-based maintenance or even serve as detection systems that assist safe driving during adverse climate conditions. In computer vision, previous work has demonstrated the effectiveness of deep learning in predicting weather conditions from outdoor images. However, training deep learning models to accurately predict weather conditions using real-world road-facing images is difficult due to: (1) the simultaneous occurrence of multiple weather conditions; (2) imbalanced occurrence of weather conditions throughout the year; and (3) road idiosyncrasies, such as road layouts, illumination, road objects etc. In this paper, we explore the use of focal loss function to force the learning process to focus on weather instances that are hard to learn with the objective to help address data imbalance. In addition, we explore the attention mechanism for pixel based dynamic weight adjustment to handle road idiosyncrasies using state-of-the-art vision transformer models. Experiments with a novel multi-label road weather dataset show that focal loss significantly increases the accuracy of computer vision approaches for imbalanced weather conditions. Furthermore, vision transformers outperforms current state-of-the-art convolutional neural networks in predicting weather conditions with a validation accuracy of 92% and F1-score of 81.22%, which is impressive considering the imbalanced nature of the dataset.

Convolutional Neural Networks (CNNs) present drawbacks for modeling geometric transformations, such as scaling and rotation, caused by the convolution operation’s locality. Deformable convolution (DCON), a mechanism that substitutes standard convolution, increasing the receptive field to capture relevant features, is a promising approach to solve this drawback and improve the robustness of CNNs. However, the optimal way to replace the standard convolution with its deformable counterpart in a CNN model is unclear. In this study, we clarify this aseveration by conducting several experiments using deformable convolutions applied in the layers that conform a small four-layer CNN model. We also use deformable convolutions on the four-layers of several ResNet CNNs with depths 18, 34, 50, and 101. The models were tested in binary balanced classes with 2D data for image classification: Cats & Dogs, EyesPACS, Spyders & Chickens, and Shapes. After this testing, we evaluated DCON in 3D data for action recognition: UCF101 and Human2 (a dataset we compiled to control movement, clothing and background). The contribution of this research lies in a guideline to use DCON. It can be summarized as follows: if DCON is used on the first layers of the proposal of model (with simple features), the computational resources expressed as the quantity of Flops will tend to increase and produce bigger misclassification than the standard CNN. However, if the DCON is used at the end layers, the quantity of Flops used in the training and testing will decrease, and the classification accuracy will improve by up to 20% about the base model. Moreover, it gains robustness when using deformable convolutions because it can adapt to the region of interest. Also, the best kernel size of the DCON is three. It showed better results than size five. In the last case, the quantity of Flops increase quadratically, but their performance does not increase significantly. With these results, we propose a guideline to use the DCON and contribute to understanding the impact of DCON on the robustness of CNNs.

The rise in both the frequency of natural disasters and the ubiquity of artificial intelligence has led to novel applications of new technologies in improving disaster response processes, such as the labor-intensive assessment of disaster damages. Assessment of residential and commercial structure damages, a precursory step to government agencies being able to provide most of their financial assistance, has benefited from aerial and satellite imagery-based computer vision models; however, limitations of using such imagery include the ground structures being obscured by clouds or smoke, as well as the lack of resolution to distinguish individual structures from others. Using a different data source, we propose a damage severity classification model using ground-level imagery, focusing on residential structures damaged by wildfires. This classifier, a Vision Transformer (ViT) model trained on over 18,000 professionally labeled images of damaged homes from the 2020-2022 California wildfires, has achieved an accuracy score of over 95%. Further, we have open sourced the training dataset–the first of its kind and scale–as well as built a publicly available web application prototype, which we demoed to and received feedback from disaster response officials, both of which further contribute to the broader literature beyond the proposed model.

Machine learning models facilitate the search for non-linear relationships when modeling hydrological processes, but they are equally effective for automation at the data preparation stage. The tasks whose automation was analyzed consisted in estimating changes in the roughness of a mountain streambed and the extent of floods from images. The Segment Anything Model (SAM) developed this year by Meta was used for this purpose. Images from many years from the Wielka Puszcza mountain stream located in the Polish Carpathians were used as the only input data. The model was not additionally trained for the described tasks. The model can be run in several modes, but the two most appropriate were used in this study. The first one is available in the form of a web application, while the second one is available in the form of a Jupyter notebook run in the Google Colab environment. Both methods do not require specialized knowledge and can be used by virtually any hydrologist. In the roughness estimation task, the average Intersection over Union (IoU) ranges from 0.55 for grass to 0.82 for shrubs/trees. Ultimately, it was possible to estimate the roughness coefficient of the mountain streambed between 0.027 and 0.059 based solely on image data. In the task of estimation of the flood extent, when selecting appropriate images, one can expect IoU at the level of at least 0.94, which seems to be an excellent result considering that SAM is a general-purpose segmentation model. It can therefore be concluded that SAM model can be a useful tool for a hydrologist.

This paper explores the application of machine learning techniques for automated object grasping. Current electronic prosthetics often require extensive training for users to gain fine motor control over the prosthetic fingers, hindering their usability and acceptance. To address this challenge, this paper proposes an automated method that leverages computer vision-based techniques and machine learning algorithms. In this study, three reinforcement learning algorithms, namely Soft Actor-Critic (SAC), Deep Q-Network (DQN), and Proximal Policy Optimization (PPO), are employed to train agents for automated grasping tasks. The results indicate that the SAC algorithm achieves the highest success rate of 99% among the three algorithms at just under 200,000 timesteps. This research also shows that object’s physical characteristics can affect the agent’s ability to learn an optimal policy. Moreover, the findings highlight the potential of the SAC algorithm in developing intelligent prosthetic hands with automatic object gripping capabilities.

This study proposes a method for image matching between infrared(IR)-RGB images using a deep learning network to estimate the deformation field. We propose a deformation field generator (DFG) that estimates the deformation field of the transformation matrix to match each pixel or IR image to the RGB image. DFG is a network that receives IR and RGB images as input; the output size is two channels and has the sample resolution as the input image. By warping the IR image through a grid-sampler that warps the image according to the value of the deformation field, we can obtain a warped IR image that matches the RGB image. Additionally, to check whether the warped IR image matched the RGB image, the masking images detecting the segmentation of objects were photographed in two images. Without directly comparing IR and RGB images, we proposed mask loss that warps the IR mask image through the deformation field and grid sampler and then compares the warped IR mask image with the RGB mask image. Mask loss solves the spatial similarity comparison problem with multi-modality images, such as IR and RGB images, by comparing the mask image with the same modality image as the mask image.

In Mexico, the incorporation of deaf people into education has been lacking since only 14% of the deaf population in the age group between 3 and 29 years access education with the support of a hearing aid. Additionally, those who have been incorporated frequently face inappropriate educational strategies which poorly develop the use of Mexican Sign Language (MSL) and therefore academical success and opportunities for insertion in the workplace are difficult. This research explores a novel mexican sign language lexicon video dataset containing the dynamical gestures most frequently used by MSL. Each gesture consists of a set of different versions of videos under uncontrolled conditions. MX-ITESO-100 data set is composed of a lexicon of 100 gestures and 5,000 videos from three participants with different grammatical elements. Additionally, the data set is evaluated in a two-step neural network model with an accuracy greater than 99%. and thus serves as a benchmark for future training of machine learning models in computer vision systems. Finally, this research provides an inclusive environment within society and organizations in particular for people with hearing impairment.

Mice are one of the frequently used animal models in science research whose behavioral characteristics can provide much valuable information in biology, neuroscience, and pharmacology. Nowadays, artificial intelligence is widely used in mice behavior analysis. Integrated AI systems such as ChatGPT and VisualGPT are already available, and we discuss the feasibility of MiceGPT to help researchers identify and classify mouse behavior more easily. We review the applications of mice behavior analysis, analyze the tasks of deep learning on these applications based on an AI pyramid, and finally summarize the AI approaches to solve these tasks. Based on these summaries, we propose three MiceGPT architectures to demonstrate the theoretical feasibility of MiceGPT.

Computer vision methods have shown to be effective in classifying garbage into recycling categories for waste processing, existing methods are costly, imprecise, and unclear. To tackle this issue, we introduce MWaste, a mobile application that uses computer vision and deep learning techniques to classify waste materials as trash, plastic, paper, metal, glass or cardboard. Its effectiveness was tested on various neural network architectures and real world images, achieving an average precision of 92% on the test set. This app can help combat climate change by enabling efficient waste processing and reducing the generation of greenhouse gases caused by incorrect waste disposal.

In this paper, our goal is to perform a virtual restoration of an ancient coin from its image. The present work is the first one to propose this problem, and it is motivated by two key promising applications. The first of these emerges from the recently recognised dependence of automatic image based coin type matching on the condition of the imaged coins; the algorithm introduced herein could be used as a pre-processing step, aimed at overcoming the aforementioned weakness. The second application concerns the utility both to professional and hobby numismatists of being able to visualise and study an ancient coin in a state closer to its original (minted) appearance. To address the conceptual problem at hand, we introduce a framework which comprises a deep learning based method using Generative Adversarial Networks, capable of learning the range of appearance variation of different semantic elements artistically depicted on coins, and a complementary algorithm used to collect, correctly label, and prepare for processing a large numbers of images (here 100,000) of ancient coins needed to facilitate the training of the aforementioned learning method. Empirical evaluation performed on a withheld subset of the data demonstrates extremely promising performance of the proposed methodology and shows that our algorithm correctly learns the spectra of appearance variation across different semantic elements, and despite the enormous variability present reconstructs the missing (damaged) detail while matching the surrounding semantic content and artistic style.

This study presents a novel approach combining computer vision with topology optimization for additive manufacturing. Using specialized photogrammetry software, we capture high-resolution images of the design domain to create precise 3D models through detailed scanning. These models are converted into STL file format and remeshed using an adaptive algorithm within COMSOL 5.3 Multiphysics, managed by a custom MATLAB 2023. This integration allows optimal mesh resolution and accuracy in analyses. We applied this method to design a concrete pillar for 3D printing, aiming for a 75% volume reduction to enhance material efficiency and stability, crucial for extraterrestrial environments. The design, scanned with a 360-degree camera array, informed the MATLAB-based topology optimization process. Combining MATLAB’s optimization algorithms with COMSOL’s meshing and finite element solver solver, we explored material-efficient configurations. The results show significant volume reduction, particularly in the central design area, optimizing material use while maintaining structural stability. The optimization algorithm demonstrated rapid, stable convergence to near-optimal solutions within about 20 iterations, highlighting the method’s efficiency and robustness in complex design optimization.

Machine learning and computer vision have proven to be valuable tools for farmers to streamline their resource utilization to lead to more sustainable and efficient agricultural production. These techniques have been applied to strawberry cultivation in the past with limited success. To build on this past work, in this study two separate sets of strawberry images, along with their associated diseases, were collected and subjected to, resizing, and augmentation. Subsequently, a combined dataset consisting of 9 classes was utilized to fine-tune three distinct pre-trained models: Vision Transformer (ViT), MobileNetV2, and ResNet18. To address the imbalanced class distribution in the dataset, each class was assigned weights to ensure nearly equal impact during the training process. To enhance the outcomes, new images were generated by removing backgrounds, reducing noise, and flipping them. The performances of ViT, MobileNetV2, and ResNet18 were compared after being selected. Customization specific to the task was applied to all three algorithms, and their performances were assessed. Throughout this experiment, none of the layers were frozen, ensuring all layers remained active during training. Attention heads were incorporated into the first 5 and last 5 layers of MobileNetV2 and ResNet18, while the architecture of ViT was modified. The results indicated accuracy factors of 98.4%, 98.1%, and 97.9% for ViT, MobileNetV2, and ResNet18, respectively. Despite the data being imbalanced, the precision, which indicates the proportion of correctly identified positive instances among all predicted positive instances, approached nearly 99% with the ViT. MobileNetV2 and ResNet18 demonstrated similar results. Overall, the analysis revealed that the Vision Transformer model exhibited superior performance in strawberry ripeness and disease classification. The inclusion of attention heads in the early layers of ResNet18 and MobileNet18, along with the inherent attention mechanism in ViT, improved the accuracy of image identification. These findings offer the potential for farmers to enhance strawberry cultivation through passive camera monitoring alone, promoting the health and well-being of the population.

In the digital age, Artificial Intelligence (AI) is emerging as a transformative force in various sectors, in-cluding medicine. This article explores the potential of AI, akin to the magical genie of Aladdin's lamp, particularly within thoracic surgery and lung cancer management. It examines AI applications like ma-chine learning and deep learning in achieving more precise diagnoses, preoperative risk assessment, and improved surgical outcomes. The challenges and advancements in AI integration, especially in computer vision and multi-modal models, are discussed alongside their impact on robotic surgery and operating room management. Despite its transformative potential, implementing AI in medicine faces challenges regarding data scarcity, interpretability issues, and ethical concerns. Collaboration between AI and medi-cal communities is essential to address these challenges and unlock the full potential of AI in revolution-izing clinical practice. This article underscores the importance of further research and interdisciplinary collaboration to ensure the safe and effective deployment of AI in real-world clinical settings.

This paper introduces an innovative approach to lipreading, leveraging a web application designed to generate subtitles for videos where the speaker's mouth is visible and a comprehensive literature review that precedes the discussion, encompassing a thorough examination of various lipreading methods employed over the past decade. Our method employs a powerful deep learning model, featuring a 3D-convolution network and bidirectional LSTM, enabling accurate sentence-level predictions based solely on visual lip movements. With an impressive accuracy of 97%, our model is trained using pre-segmented lips regions, transformed into animated GIFs for effective pre-training. This work stands as a significant contribution to the evolving landscape of lipreading research, offering a practical and accurate solution for real-world applications.

The point cloud is a form of three-dimensional data that comprises various detailed features at multiple scales. Due to this characteristic and its irregularity, point cloud analysis based on deep learning is challenging. While previous works utilize the sampling-grouping operation of PointNet++ for feature description and then explore geometry by means of sophisticated feature extractors or deep networks, such operations fail to describe multi-scale features effectively. Additionally, these techniques have led to performance saturation. And it is difficult for standard MLPs to directly "mine" point cloud. To address above problems, we propose the Detail Activation (DA) module, which encodes data based on Fourier transform after sampling and grouping. We activate the channels at different frequency levels from low to high in the DA module to gradually recover finer point cloud details. As training progresses, the proposed Point-MDA can uncover local and global geometries of point cloud progressively. Our experiments show that Point-MDA achieves superior classification accuracy, outperforming PointNet++ by 3.3% and 7.9% in terms of overall accuracy on the ModelNet40 and ScanObjectNN dataset, respectively. Furthermore, it accomplishes this without employing complicated operations, while exploring the full potential of PointNet++.

In the realm of computer vision, semantic segmentation is the task of recognizing objects in images at the pixel level. This is done by performing a classification of each pixel. The task is complex and requires sophisticated skills and knowledge about the context to identify objects’ boundaries. The importance of semantic segmentation in many domains is undisputed. In medical diagnostics, it simplifies the early detection of pathologies, thus mitigating the possible consequences. In this work, we provide a review of the literature on deep ensemble learning models for polyp segmentation and we develop new ensembles based on convolutional neural networks and transformers. The development of an effective ensemble entails ensuring diversity between its components. To this end, we combine different models (HarDNet-MSEG, Polyp-PVT, and HSNet) trained with different data augmentation techniques, optimization methods, and learning rates, which we experimentally demonstrate to be useful to form a better ensemble. Most importantly, we introduce a new method to obtain the segmentation mask which is more suitable for combining transformers in an ensemble. In our extensive experimental evaluation, the proposed ensembles exhibit state-of-the-art performance.

A single paragraph of about 200 words maximum. For research articles, abstracts should give a pertinent overview of the work. We strongly encourage authors to use the following style of structured abstracts, but without headings: (1) Background: place the question addressed in a broad context and highlight the purpose of the study; (2) Methods: describe briefly the main methods or treatments applied; (3) Results: summarize the article’s main findings; (4) Conclusions: indicate the main conclusions or interpretations. The abstract should be an objective representation of the article, it must not contain results which are not presented and substantiated in the main text and should not exaggerate the main conclusions.

The use of Attitude and Heading Reference Systems (AHRS) for orientation estimation is now common practice in a wide range of applications, e.g., robotics and human motion tracking, aerial vehicles and aerospace, gaming and virtual reality, indoor pedestrian navigation and maritime navigation. The integration of the high-rate measurements can provide very accurate estimates, but these can suffer from errors accumulation due to the sensors drift over longer time scales. To overcome this issue, inertial sensors are typically combined with additional sensors and techniques. As an example, camera-based solutions have drawn a large attention by the community, thanks to their low-costs and easy hardware setup; moreover, impressive results have been demonstrated in the context of Deep Learning. This work presents the preliminary results obtained by DOES , a supportive Deep Learning method specifically designed for maritime navigation, which aims at improving the roll and pitch estimations obtained by common AHRS. DOES recovers these estimations through the analysis of the frames acquired by a low-cost camera pointing the horizon at sea. The training has been performed on the novel ROPIS dataset, presented in the context of this work, acquired using the FrameWO application developed for the scope. Promising results encourage to test other network backbones and to further expand the dataset, improving the accuracy of the results and the range of applications of the method.

Auto checkout has received more and more attention in recent years and this system automatically generates a shopping bill by identifying the picture of the products purchased by the customers. However, the system is challenged by the domain adaptation problem, where each image of the training set contains only one commodity, whereas the test set is a collection of multiple commodities. The existing solution to this problem is to resynthesize the training images to enhance the training set. Then the composite images are rendered using CycleGAN to make the image distribution of the training set and the test set more similar. However, we find that the detection boxes given by the ground truth of the common dataset contain a large part of the background area, the area will affect the training process as noise. To solve this problem, we propose a mask data priming method. Specifically, we redo the large scale Retail Product Checkout (RPC) dataset and add segmentation annotation information to each item in the training set image based on the original dataset using pixel-level annotation. Secondly, a new network structure is proposed in which we train the network using joint learning of detectors and counters, and fine-tune the detection network by filtering out suitable images from the test set. Experiments on the RPC dataset have shown that our method yields better results. we used an approach that reached 81.87% compared to 56.68% for the baseline approach which demonstrates that pixel-level information helps to improve the detection results of the network.

Spike shape and morphometric characteristics are among the key characteristics of cultivated cereals associated with their productivity. Identification of the genes controlling these traits requires morphometric data harvesting and analysis for numerous plants, which is automatable using technologies of digital image analysis. A method for wheat spike morphometry utilizing 2D image analysis is proposed. Digital images are acquired in two variants: a spike on a table (one projection) or fixed with a clip (four projections). The method identifies spike and awns in the image and estimates their quantitative characteristics (area in image, length, width, circularity, etc.). Models of sections, quadrilaterals, and radial model are proposed for describing spike shape. Parameters of these models are used to predict spike shape type (spelt, normal, or compact) by machine learning. The mean error in spike density prediction for the images in one projection is 4.61 (~18%) versus 3.33 (~13%) for the parameters obtained using four projections. F1 measure in automated spike classification into three types is 0.78 using logistic regression (one projection) and 0.85 using random forest method (four projections). The proposed method is implemented in Java; examples of images and user guide are available at http://wheatdb.org/werecognizer.

In high-risk railway construction, personal protective equipment monitoring is critical but challenging due to small and frequently obstructed targets. We propose YOLO-EA, an innovative model that enhances safety measure detection by integrating ECA into its backbone's convolutional layers, improving discernment of minuscule objects like hardhats. YOLO-EA further refines target recognition under occlusion by replacing GIoU with EIoU loss. YOLO-EA's effectiveness was empirically substantiated using a dataset derived from real-world railway construction site surveillance footage. It outperforms YOLOv5, achieving 98.9% precision and 94.7% recall, up 2.5% and 0.5% respectively, while maintaining real-time performance at 70.774 fps. This highly efficient and precise YOLO-EA holds great promise for practical application in intricate construction scenarios, enforcing stringent safety compliance during complex railway construction projects.

Through the refinement of approaches to applying visual augmentation to virtual environments, the possibilities of utilising this process in robotics continues to be explored. However these endeavours have only yielded a collaborative relationship between the two technologies, where the augmented reality (AR) component is separate from the robot functionality. Regarding telepresence robots, research has shown that their perceived usefulness wanes in comparison to that of video conferencing tools as they do not offer a level of immersion and modified visual experience to justify their use. Therefore this paper contributes a novel virtually augmented robot (VAR) model that embeds AR functionality into telepresence, through the use of augmentation and facial recognition sub-models and algorithms. Experimental results show that using pulse width modulation (PWM), duty cycles, digital signals and accelerometer angles provides an approach for stabilising robot motion and the transference of human to robot head movement. The results also show the benefit of having a robot model with intrinsic AR functionality, as it offers a greater level of immersion and visual control of an environment.

The overpopulation of feral pigeons in Hong Kong has significantly disrupted the urban ecosystem, highlighting the urgent need for effective strategies to control their population. In general, control measures should be implemented and re-evaluated periodically following accurate estimations of feral pigeon population in the concerned regions, which, however, is very difficult in urban envi-ronments due to the concealment and mobility of pigeons within complex building structures. With the advances in deep learning, computer vision can be a promising tool for pigeon monitoring and population estimation but has not been well investigated so far. Therefore, we propose an improved deep learning model based on Mask-RCNN (Swin-Mask R-CNN) for feral pigeon detection using computer vision techniques. Specifically, our model consists of a Swin transformer network (STN) as the backbone, a feature pyramid network (FPN) as the neck, and three decoupled detection heads. The STN is utilized to extract deep feature information of feral pigeons through local and cross-window attention mechanisms. The FPN is employed to fuse multi-scale features and enhance the multi-scale learning ability. Heads in the three branches are responsible for classification, pre-dicting best bounding boxes, and segmentation of feral pigeons, respectively. During the prediction phase, a Slicing Aided Hyper Inference (SAHI) tool is employed to zoom in on the feature infor-mation of small feral pigeon targets, and the segmentation head is frozen to expedite inference of large images. Experiments were conducted on feral pigeon dataset to evaluate model performance. The results reveal that our model is well-suited for detecting small targets in high-resolution images and achieves excellent recognition performance for feral pigeons with a mAP (mean average pre-cision) and an AP50 (average precision at 50% intersection over union) of 0.74 and 0.93, respectively. For small target feral pigeons, AP50 in small scale (AP50s) improved by 10% as compared to the Mask R-CNN (AP50s of 0.75), demonstrating its potential for dynamic pigeon detection and population estimation in the future.

Track fasteners play a pivotal role in infrastructure inspection for high-speed rail. Yet, images taken by drones often capture shadows cast by electrical towers flanking the high-speed rail tracks. These shadows can hinder the visibility of the track fasteners, thereby impacting detection efficiency and accuracy considerably. The present paper introduces an end-to-end shadow removal algorithm, rooted in generative adversarial network training. The comprehensive network framework is segmented into three sub-networks: pseudo-mask generation, shadow removal, and result refinement. We have integrated a Fourier convolutional residual module to bolster the feature extraction capability of the generator network. This integration ensures the network retains a global receptive field, even in its more superficial layers. By employing an overall weighted loss function, we enhance the quality of the images produced without shadows. Further, a perceptual loss function has been incorporated to retain the structural information of objects, setting the stage for subsequent defect detection. Our results highlight that Pse-ShadowNet adeptly eradicates fastener shadows while maintaining vital visual features, including object position, structure, texture, edges, and other key visual elements. Consequently, the reconstructed images are detailed and showcase superior image quality.

Human Action Recognition is widely used in multiple fields to recognize activities and extract spatial and temporal information. This paper focuses on analyzing various methods as well as provides extensive knowledge of the foundational concepts of the HAR. A dataset is also very crucial for any research study so we discussed the popular dataset and its features in this paper.

With the advent of deep learning, significant progress has been made in low-light image enhancement methods. However, deep learning requires enormous paired training data, which is challenging to capture in real-world scenarios.To address this limitation, this paper presents a novel unsupervised low-light image enhancement method, which first introduces the frequency domain features of images in low-light image enhancement tasks. Our work is inspired by imagining a digital image as a spatially varying metaphoric “field of light”, then subjecting the influence of physical processes such as diffraction and coherent detection back onto the original image space via a frequency-domain to spatial-domain transformation (inverse Fourier transform). However, the mathematical model created by this physical process still requires complex manual tuning of the parameters for different scene conditions to achieve the best adjustment. Therefore, we proposed a dual-branch convolution network to estimate pixel-wise and high-order spatial interactions for dynamic range adjustment of the frequency feature of the given low-light image. Guided by the frequency feature from the “field of light” and parameter estimation networks, our method enables dynamic enhancement of low-light images. Extensive experiments have shown that our method performs well compared to state-of-the-art unsupervised, and its performance approximates the level of the state-of-the-art supervised methods qualitatively and quantitatively. At the same time, the light network structure design allows the proposed method to have an extremely fast inference speed(nearly 150 FPS on an NVIDIA 3090 Ti GPU for an image of size 600×400×3 ). Furthermore, the potential benefits of our method to object detection in the dark are discussed.

Objective tools to track medication adherence are lacking. A tool to monitor pill intake that can be implemented in mHealth apps without the need for additional devices was developed. We propose a pill intake detection tool that uses digital image processing to analyze images of a blister to detect the presence of pills. The tool uses the circular Hough transform as a feature extraction technique and is therefore primarily useful for the detection of pills with a round shape. This pill detection tool is composed of two steps. First, the registration of a full blister and storing of reference values in a local database. Second, the detection and classification of taken and remaining pills in similar blisters, to determine the actual number of untaken pills. In the registration of round pills in full blisters, 100% of pills in gray blisters or blisters with a transparent cover were successfully detected. In counting of untaken pills in partially opened blisters, 95.2% of remaining and 95.1% of taken pills were detected in gray blisters, while 88.2% of remaining and 80.8% of taken pills were detected in blisters with a transparent cover. The proposed tool provides promising results for the detection of round pills. However, the classification of taken and remaining pills need to be further improved, in particular for the detection of pills with non-oval shapes.

Long-term continuous time series of SO2 emissions are considered critical elements of both volcano monitoring and basic research into processes within magmatic systems. One highly successful framework for computing these fluxes involves reconstructing a representative time-averaged SO2 plume from which to estimate the SO2 source flux. Previous methods within this framework have used ancillary wind datasets from reanalysis or numerical weather prediction (NWP) to construct the mean plume and then again as a constrained parameter in the fitting. Additionally, traditional SO2 datasets from ultraviolet (UV) sensors lack altitude information which must be assumed to correctly calibrate the SO2 data and to capture the appropriate NWP wind level which can be a significant source of error. We have made novel modifications to this framework which do not rely on prior knowledge of the winds and therefore do not inherit errors associated with NWP winds. To perform the plume rotation, we modify a rudimentary computer vision algorithm designed for object detection in medical imaging to detect plume-like objects in gridded SO2 data. We then fit a solution to the general time-averaged dispersion of SO2 from a point source. We demonstrate these techniques using SO2 data generated by a newly developed probabilistic layer height and column loading algorithm designed for the Cross-track Infrared Sounder (CrIS), a hyperspectral infrared sensor aboard the Joint Polar Satellite System’s Suomi-NPP and NOAA-20 satellites. This SO2 data source is best suited to flux estimates at high-latitude volcanoes and at low-latitude, but high-altitude volcanoes. Of particular importance, IR SO2 data can fill an important data gap in the UV-based record: estimating SO2 emissions from high-latitude volcanoes through the polar winters when there is insufficient solar backscatter for UV sensors to be used.

The article describes a classification solution for amber stones. The problem of classifying amber is known for a long time among jewelers and artisans of amber art. Existing solutions can classify amber pieces according to color, but a need to classify by shape and texture is not satisfied up to now. The proposed solution is capable of classifying the gemstones according to a shape. Amber can be considered as a specific object since the form is difficult to define unambiguously. Data for amber experiments was gathered from amber art craftsmen. In the proposed solution amber form can be classified into 10 different classes (7 classes chosen during the experiment).

Ancient numismatics, that is, the study of ancient currencies (predominantly coins), is an interesting domain for the application of computer vision and machine learning, and has been receiving an increasing amount of attention in recent years. Notwithstanding the number of articles published on the topic, the variety of different methodological approaches described, and the mounting realisation that the relevant problems in the field are most challenging indeed, all research to date has entirely ignored one specific, readily accessible modality: colour. Invariably, colour is discarded and images of coins treated as being greyscale. The present article is the first one to question this decision (and indeed, it is a decision). We discuss the reasons behind the said choice, present a case why it ought to be reexamined, and in turn investigate the issue for the first time in the published literature. Specifically, we propose two new colour-based representations specifically designed with the aim of being applied to ancient coin analysis, and argue why it is sensible to employ them in the first stages of the classification process as a means of drastically reducing the initially enormous number of classes involved in type matching ancient coins (tens of thousands, just for Ancient Roman Imperial coins). Furthermore, we introduce a new data set collected with the specific aim of denomination-based categorisation of ancient coins, where we hypothesised colour could be of potential use, and evaluate the proposed representations. Lastly, we report surprisingly successful performances which goes further than confirming our hypothesis—rather, they convincingly demonstrate a much higher relevant information content carried by colour than even we expected. Thus we trust that our findings will be noted by others in the field and that more attention and further research will be devoted to the use of colour in automatic ancient coin analysis.

Image captioning aims at generating meaningful verbal descriptions of a digital image. Our paper focuses on the classic encoder-decoder deep learning model that consists of several components – sub-networks, each performing a separate task that, combined, form an effective caption generator. We investigate image feature extractors, recurrent neural networks, word embedding models, and word generation layers and discuss how each component influences the captioning model’s overall performance. Our experiments are performed on the MS COCO 2014 dataset. The results help design efficient models with optimal combinations of their components.

Image clustering and classification are fundamental tasks in computer vision, critical for applications overlap image retrieval, object recognition, and image classification. However, the robustness of clustering algorithms against adversarial attacks remains interesting topic. In this paper, we investigate how adversarial attacks on image classification algorithms impact Image Clustering, similarity obtained using the Dot Product, KNN, HNSW algorithms and model Gradient-Weighted Class Activation Mapping (Grad-CAM). In our work was proposed a targeted study of the impact of adversarial attacks on the clustering ability of ResNet50 under various adversarial scenarios. Was used ResNet50 as the basis for the experiments, a widely used architecture known for its effectiveness in image classification. This network was subjected to various adversarial attacks in order to understand how these perturbations affect its clustering capabilities. By thoroughly examining the resultant clustering outcomes under different attack scenarios, we aim to uncover vulnerabilities and nuances inherent in clustering algorithms and similarity metrics when confronted with adversarial input.

Detecting cracks in structural health monitoring is crucial for ensuring infrastructure safety and longevity. Using drones to obtain crack images and automate processing can improve the efficiency of crack detection. To address the challenges posed by the limited computing resources of edge devices in practical applications, we propose CrackScopeNet, a lightweight segmentation network model that simultaneously considers local and global crack features, suitable for deployment on drone platforms with limited computational power and memory. CrackScopeNet features a multi-scale branch to improve sensitivity to cracks of varying sizes without substantial computational overhead and a stripe-wise context attention mechanism to enhance the capture of long-range contextual information, mitigating the interference from complex backgrounds. Experimental results on CrackSeg9k dataset demonstrate that CrackScopeNet leads to a significant improvement in prediction performance, with the highest mean intersection over union (mIoU) scores reaching 82.12%, while maintaining a lightweight architecture with only 1.05M parameters and 1.58G floating point operations (FLOPs). Besides, CrackScopeNet excels in inference speed on edge devices without GPU because of its low FLOPs. CrackScopeNet contributes to the development of efficient and effective crack segmentation networks, suitable for practical structural health monitoring applications using drone platforms.

In the realm of agribusiness, transformative shifts are underway, propelled by the growing demands and expanding scales of grain production. This evolution calls for a critical reevaluation of the existing paradigms in coffee production and marketing paradigms, with a specific focus on integrating Artificial Intelligence (AI). This work aims to review, synthesize, and summarize the available data regarding how Machine Learning (ML) has been used to detect and classify characteristics in coffee beans and leaves. For this purpose, a comprehensive literature review of the most significant research contributions describing the application of AI for advanced classification techniques in coffee agriculture has been carried out. Our analysis suggests that implementing AI technologies allows the classification of coffee, encompassing various attributes such as maturity, roast intensity, disease identification, flavor profiles, and overall quality. More largely, this technological advancement holds the potential to revolutionize coffee farming by providing producers and agricultural specialists with sophisticated tools to enhance production efficiency, minimize costs, and improve the accuracy and confidence of their decision-making processes.

The exponential growth of maritime traffic has brought with it a multitude of business opportunities, while at the same time posing unprecedented challenges in terms of control and surveillance. Traditional methods struggle to cope with the sheer volume and complexity of maritime activities, which calls for innovative solutions. This study presents a comparative analysis of three prominent computing object detection models applied to maritime surveillance. By evaluating their performance metrics in dynamic maritime environments, this study provides information on the most suitable model for effective maritime monitoring applications, attaining an impressive 91\% mean Average Precision (mAP) using the YOLOv5n model.

Within environmental perception, automatic navigation and object detection, computer vision is a crucial and demanding field with many applications in modern industries, such as multi-target long-term visual tracking in automated production, defect detection, and driverless robotic vehicles with sensing ability. The performance of computer vision has greatly improved recently thanks to developments in deep learning algorithms and hardware computing capabilities, which has spawned the creation of a large number of related applications. This paper presents the results of a detailed review of over 50 papers published over the course of two decades (1999–2024), with a primary focus on the technical advancement of computer vision. To elucidate the foundational principles, an examination of typical visual algorithms based on traditional correlation filtering was initially conducted. Subsequently, a comprehensive overview of the most recent advancements in deep learning-based computer vision techniques was compiled. Furthermore, a comparative analysis between conventional and novel algorithms was undertaken to discuss the future trends and directions of computer vision. Lastly, the feasibility of employing visual SLAM (Simultaneous Localization and Mapping) algorithms in the context of autonomous vehicles with sensing capability was explored. Furthermore, we explored a thorough and efficient architecture for utilizing these techniques in autonomous robotic vehicles into unmanned green and low-carbon smart factories, underlining novel advances as well as potential prospects for future research.

Reconstructing a 3D model of a scene from a set of multiple views poses a significant challenge in the field of computer vision. The advent of NeRF has marked a major breakthrough in this domain. However, there is a need to extend the capabilities of the NeRF model to enable editing tasks such as relighting and deformation, as well as to enhance its training efficiency. Neural reflectance fields introduce the concept of the reflectance equation into the NeRF framework to achieve relighting of the NeRF model. We leverage the TensoRF approach, which incorporates tensor decomposition and employs multiple tensors to store features, to expedite the training process of the NeRF model. Our novel method, called StensorR, combines the reflectance equation and tensor decomposition within the radiation field model framework. Differing from previous approaches, we employ a single tensor to store scene features and render the surface color of our scene using a simplified reflectance equation. This approach accelerates model convergence and enables relighting of the NeRF model. Experimental results demonstrate that our method achieves a 50% faster convergence rate compared to existing relighting radiation field models, while successfully enabling relighting and improving the quality of synthesized images from new viewpoints.

Convolutional neural network (CNN), a class of artificial neural network (ANN) is attracting interests of researchers in all research domain. CNN was invented for computer vision. They have also shown to be useful for semantic parsing, sentence modeling and other natural language processing related tasks. Here in this paper we discuss the basics of CNN models and their scope to provide a reference/baseline to the researchers interested in using CNN models in their research.

The contamination on the insulators may increase its surface conductivity and, as a consequence, electrical discharges occur more frequently, which can lead to interruptions in the power supply. To maintain reliability in the electrical distribution power system, components that have lost their insulating properties must be replaced. Identifying the components that need maintenance, is a difficult task as there are several levels of contamination that are hardly noticed during inspections. To improve the quality of inspections, this paper proposes to use the k-nearest neighbours (k-NN) to classify the levels of insulator contamination, based on the image of insulators at various levels of contamination simulated in the laboratory. Using computer vision features such as mean, variance, asymmetry, kurtosis, energy, and entropy are used for training the k-NN. To assess the robustness of the proposed approach, statistical analysis and a comparative assessment with well-consolidated algorithms such as decision tree, ensemble subspace, and support vector machine models are presented. The k-NN showed results of up to 85.17 % accuracy using the k-fold cross-validation method, with an average accuracy higher than 82 % for multi-classification of the contamination of the insulators, being superior to the compared models.

We have an unprecedented ability to map the Earth’s surface as deep learning technologies are applied to an abundance of high-frequency Earth observation data. Simple, free, and effective methods are needed to enable a variety of stakeholders to use these tools to improve scientific knowledge and decision making. Here we present a trained U-Net model that can map and delineate ground mounted solar arrays using publicly available Sentinel-2 imagery, and that requires minimal data pre-processing and no feature engineering. By using label overloading and image augmentation during training, the model is robust to temporal and spatial variation in imagery. The trained model achieved a precision and recall of 91.5% each and an intersection over union of 84.3% on independent validation data from two distinct geographies. This generalizability in space and time makes the model useful for repeatedly mapping solar arrays. We use this model to delineate all ground mounted solar arrays in North Carolina and the Chesapeake Bay watershed to illustrate how these methods can be used to quickly and easily produce accurate maps of solar infrastructure.

This paper presents a state machine based architecture which enhances flexibility and reusability of industrial robots, more concretely dual-arm multisensor robots. The proposed architecture, in addition to allowing absolute control of the execution, eases the programming of new applications by increasing the reusability of the developed modules. Through an easy-to-use graphical user interface, operators are able to create, modify, reuse and maintain industrial processes increasing the flexibility of the cell. Moreover, the proposed approach is applied in a real use case in order to demonstrate its capabilities and feasibility in industrial environments. A comparative analysis is presented for evaluating presented approach versus traditional robot programming techniques.

Tracking and early identification of suspected cases are essential to control and prevent potential COVID-19 outbreaks. One of the most popular techniques used to track this disease is the use of Infrared cameras to identify individuals with elevated body temperatures. However, they are limited by their inability to be implemented in open public settings such as public parks or even outdoor recreational centers. This limits their ability to effectively track possible COVID-19 patients as open public recreational places such as parks, concert venues and other public venues are hotspots for the spreading of the virus. Other technological solutions such as thermal scanners require an individual to perform the actual testing as they are not individual standalone technologies. This method of testing can potentially cause the transmission of the virus between the tester and the individual getting tested. As can be seen, an alternative solution is essential to solving this issue. In this study, we aim to present the system, design and potential scope of a non-invasive system that can diagnose and identify potential COVID-19 patients using thermal and optical images of the individual using drone technology. The proposed system (COVIDRONE) combines multi-modal machine intelligence, computer vision and real-time monitoring to enable scalable monitoring. The system will also involve the use of machine learning algorithms for better and more accurate diagnosis. We envisage that development of such technologies may help in developing technological solutions to combat infectious disease threats in the future pandemics.

Traditional computer vision techniques aim to extract meaningful information from images but often depend on manual feature engineering, making it difficult to handle complex real-world scenarios. Fractional Calculus (FC), which extends derivatives to non-integer orders, provides a flexible way to model systems with memory effects and long-term dependencies, making it a powerful tool for capturing fractional rates of variation. Recently, Neural Networks (NNs) have demonstrated remarkable capabilities in learning complex patterns directly from raw data, automating computer vision tasks and enhancing performance. Therefore, the use of fractional calculus in neural network-based computer vision is a powerful method to address existing challenges by effectively capturing complex spatial and temporal relationships in images and videos. This paper presents a survey of Fractional Calculus Neural Network-based (FC NN-based) computer vision techniques for denoising, enhancement, object detection, segmentation, restoration, and NN compression. The survey compiles existing FFC NN-based approaches, elucidates underlying concepts, and identifies open questions and research directions. By leveraging FC's properties, FC NN-based approaches offer a novel way to improve the robustness and efficiency of computer vision systems.

Over the last few decades, advancements in astrophysics have been closely linked to the development of powerful machine-learning models that can accurately classify celestial bodies. At the same time, however, many astronomical datasets are filled with new features collected by increasingly powerful telescopes. These features can cause overfitting, clouding predictive abilities, and dampening the ability of many models to classify images. Therefore, motivation to design more efficient models has skyrocketed—aiming to optimize for lower run times and high accuracies, even with fewer provided features. In our project, we seek to optimize a convolutional neural network model using a technique known as wavelet analysis. This technique allows us to pick the key features of an astronomical image and accentuate niche details, saving time and boosting accuracy. We applied it to the MiraBest Dataset, a dataset compiled from the FIRST sky survey using a virtual telescope. In the end, after training our neural network on the original images and the five filters (approximation, horizontal, vertical, diagonal, and combined), we found that with fewer features and less overfitting, the vertical Daubechies-family wavelet filter outperformed the original runs with the unaltered images by over 10%. Our findings suggest that wavelet analysis can help harvest the most valuable features in images of celestial bodies–leading to enhanced predictions in astronomical applications and perhaps bolstering modern astrophysical theory.

The production of cheese, a beloved culinary delight worldwide, faces challenges in maintaining consistent product quality and operational efficiency. One crucial stage in this process is determining the precise cutting time during curd formation, which significantly impacts the quality of the cheese. Misjudging this timing can lead to the production of inferior products, harming a company's reputation and revenue. Conventional methods often fall short of accurately assessing variations in coagulation conditions due to the inherent potential for human error. To address this issue, we propose an anomaly detection-based approach. In this approach, we treat the class representing curd formation as the anomaly to be identified. Our proposed solution involves utilizing a one-class, fully convolutional data description network, which we compared against several state-of-the-art methods to detect deviations from the standard coagulation patterns. Encouragingly, our results show F1 scores of up to 0.92, indicating the effectiveness of our approach.

Efficient processing of end-of-life lithium-ion batteries of electric vehicles is important and a pressing challenge for a circular economy. Regardless of whether the processing strategy is recycling, repurposing or remanufacturing, the first processing step would usually involve disassembly. As battery disassembly is a dangerous task, efforts have been made to robotise it. In this paper, a robotic disassembly platform using four industrial robots is proposed to automate the non-destructive disassembly of a plug-in hybrid electric vehicle battery pack into modules. The work was conducted as a case study to demonstrate the concept of autonomous disassembly of an electric vehicle battery pack. A two-step object localisation method based on visual information is used to overcome positional uncertainties from different sources and is validated by experiments. Also, the unscrewing system is highlighted, and its functions, such as handling untightened fasteners, loosening jammed screws, and changing the nutrunner adapters with square drives, are detailed. Furthermore, the time required for each operation is compared with that taken by human operators. Finally, the limitations of the platform are reported and future research directions are suggested.

In the age of social media, images come from unreliable sources that may change their content to fit a narrative. Automated detection of forged images is a complex task, especially considering recent technological advances in image manipulation software. However, in the existing literature, it is possible to identify two general approaches adopted by detection methods, active and passive. Active techniques preemptively act on an image by inserting structures before any manipulation is made that can verify its authenticity. In contrast, passive methods analyze the content of an image in search of traces of manipulation. In this way, this research proposes a novel approach to image manipulation detection, combining two passive methods through neural networks, creating a generalist approach capable of detecting with greater accuracy than the methods that compose it. Furthermore, this work used a combination of four datasets available in the literature for training and evaluation. After training, the merged approach obtained an accuracy of 89.59% in the set of validation images, significantly higher than the model trained with only unaltered images, which obtained 78.64%, and the two other models trained using images with a feature selection method applied to enhance inconsistencies that obtained 68.02% for Error-Level Analysis images and 50.70% for the method using Discrete Wavelet Transform. In addition to the performance improvement, the proposed approach’s accuracy variation was lower than the other models.

Computer vision is a powerful technology that has enabled solutions in various fields by analyzing visual attributes in images. One field that has taken advantage of computer vision is agricultural automation, which promotes high-quality crop production. The nutritional status of a crop is a crucial factor in determining its productivity. This status is mediated by approximately 14 chemical elements acquired by the plant, and their determination plays a pivotal role in farm management. To address the timely identification of nutritional disorders, this study focuses on the classification of three levels of phosphorus deficiencies through individual leaf analysis. The methodological steps include: (1) generating a database with laboratory-grown maize plants that were induced to total phosphorus deficiency, medium deficiency, and total nutrition, using different capture devices; (2) processing the images with state-of-the-art transfer learning architectures (i.e. VGG16, ResNet50, GoogLeNet, DenseNet201, and MobileNetV2); and (3) evaluating the classification performance of the models using the created database. The results show that the VGG16 model achieves superior performance, with 98% classification accuracy. However, the other studied architectures also demonstrate competitive performance and are considered state-of-the-art automatic leaf deficiency detection tools. The proposed method can be a starting point to fine-tune machine vision-based solutions tailored for real-time monitoring of crop nutritional status.

Iron ore processing involves critical steps that affect the quality of the final product. Determining the diameter of the pellets is the necessary initial step for volume measurement and, ultimately, for porosity and bulk density measurement, crucial characteristics for optimizing the burning process in the blast furnace. Traditional measurement methods using mercury present issues related to operator safety and environmental preservation, while the practices described in ISO 4698 standard require lengthy preparation and execution time. In light of environmental needs, operator safety, and the time consumed in tests, implementing a new method based on digital image processing and convolutional neural networks for measuring the diameter of burnt iron ore pellets is proposed. The methodology involves capturing images of the pellets using a high-resolution camera and utilizing digital image processing and neural networks capable of performing pixel-by-pixel object segmentation in the images, providing precise information about the pellets to calculate their volume automatically. The results were compared with those obtained using traditional methods, evaluating their conformity with the ISO 4698 standard. In conclusion, this study offers a new approach to measuring the volume of burnt iron ore pellets, providing accurate and reliable results in compliance with safety and environmental preservation standards.

Quality control is a critical component in industrial manufacturing, directly influencing efficiency, product reliability, and ultimately, customer satisfaction. In the dynamic environment of industrial manufacturing, traditional methods of inspection may not adequately meet the evolving complexity, necessitating innovative approaches to bolster precision and productivity. In this study, we explore the application of mixed reality (MR) technology for real-time quality control in the assembly process. Our methodology involved the integration of smart glasses with a server-based image recognition system, designed to conduct real-time component analysis. The innovative aspect of our study lies in the harmonization of MR and computer vision algorithms, providing immediate visual feedback to inspectors and thereby improving the speed and accuracy of defect detection. YOLOv8 have been adopted in this study for detection object model. The project implementation occurred in a controlled environment to enable a comprehensive evaluation of the system functionality, the identification of possible problems and improvements in the system performance. The results indicated the viability of mixed reality as a powerful tool for enhancing traditional inspection processes. The fusion of MR and computer vision offers possibilities for future advancements in industrial quality control, paving the way for more efficient and reliable manufacturing ecosystems.

Studying and understanding the behavior of people and vehicles on public roads can be of utmost importance for supporting the activities of many institutional stakeholders. It may allow automated supervision of the ongoing situation in a given place, with warnings or alarms raised in case of anomalies. It may be used to plan their interventions on road and town organization. It may provide them with advanced support to decision-making. The number of involved entities and places to manage makes it infeasible to manually handle all the traffic-related tasks. Moreover, the complexity of the tasks to be carried out requires the adoption of advanced approaches. Many AI solutions are nowadays mature to support these requirements. In some cases, the motivations and objectives of traffic management require the AI outcomes to be understandable, interpretable and explainable. In this paper, we propose TrAnSIT (TRaffic ANalysis Supervision and Interpretation Tool), an AI-based framework that combines several modules, each aimed at tackling a specific traffic-related task, so as to cover a wide landscape of traffic-related issues, from overall urban or suburban traffic management to surveying specific road segments that fall under the scope of one camera. Most of these modules are based on AI techniques that support a human-level understanding of the outcomes.

The abundance of information in academic articles, reports, and studies can make it challenging for researchers to gain insights from the existing literature. To address this issue, there is a growing demand for tools that can help researchers effectively parse and analyze large volumes of data. One such tool is DataDiscoveryLab, a software system that utilizes computer vision algorithms and NLP techniques to parse academic articles into text and figures, creating three separate databases. These databases allow researchers to quickly identify articles that may be relevant to their research questions, gain a deeper understanding of the research presented, and analyze visual data. The integration of article mining and computer vision in the DataDiscoveryLab software system provides researchers with a powerful tool for navigating the vast amount of scientific literature available today. Yet, as we will discuss in the latter papers these databases’ purpose is to create a bridge between researchers’ data and practically unlimited scientific publications. Yet, in this article, we will discuss how we plan to do that, and our efforts on integrating deep learning modes. After all, unlike already existing AI models, DataDiscoveryLab can be their combination and the first Generative AI in academia that can encompass every part of the natural sciences.

As machine learning and computer vision techniques and methods continue to advance, the collection of naturalistic traffic data from video feeds is becoming more and more feasible. That is especially true for the case of bicycles, for which the collection of naturalistic data is not achievable in the traditional vehicle approach. This study describes a research effort that aims to extract naturalistic cycling data from a video dataset for use in safety and mobility applications. The used videos come from a dataset collected in a previous Virginia Tech Transportation Institute study in collaboration with SPIN in which continuous video data at a non-signalized intersection on the Virginia Tech campus was recorded. The research team applied computer vision and machine learning techniques to develop a comprehensive framework for the extraction of naturalistic cycling trajectories. In total, this study resulted in the collection and classification of 619 bicycle trajectories based on their type of interactions with other road users. The results confirm the success of the proposed methodology in relation to extracting the locations, speeds, and accelerations of the bicycles at a high level of precision. Furthermore, preliminary insights into the acceleration and speed behavior of bicyclists around motorists are determined.

The study purpose was to train and validate a deep-learning approach to detect micro-scale streetscape features related to pedestrian physical activity. This work innovates by combining computer vision techniques with Google Street View (GSV) images to overcome impediments to conducting audits (e.g., time, safety, and expert labor cost). The EfficientNETB5 architecture was used to build deep-learning models for eight micro-scale features guided by the Microscale Audit of Pedestrian Streetscapes-Mini tool: sidewalks, sidewalk buffers, curb cuts, zebra and line crosswalks, walk signals, bike symbols, and streetlights. We used a train--correct loop, whereby images were trained on a training dataset, evaluated using a separate validation dataset, and trained further until acceptable performance metrics were achieved. Further, we used trained models to audit participant (N=512) neighborhoods in the WalkIT Arizona trial. Correlations were explored between micro-scale features and GIS-measured- and participant reported-macro-scale walkability. Classifier precision, recall, and overall accuracy were all >84%. Total micro-scale was associated with overall macro-scale walkability (r=0.300,p<.001). Positive associations were found between model-detected and self-reported sidewalks (r=0.41,p<.001) and sidewalk buffers (r=0.26,p<.001). Computer vision model results suggest an alternative to trained human raters, allowing for audits of hundreds or thousands of neighborhoods for population surveillance or hypothesis testing.

Animal welfare is not only an ethically important consideration in good animal husbandry, but can also have a significant effect on an animal’s productivity. The aim of this paper is to show that a reduction in animal welfare, in the form of increased stress, can be identified in pigs from frontal images of the animals. We train a Convolutional Neural Network (CNN) using a leave-one-out design and show that it is able to discriminate between stressed and unstressed pigs with an accuracy of >90% in unseen animals. Grad-CAM is used to identify the animal regions used, and these support those used in manual assessments such as the Pig Grimace Scale. This innovative work paves the way for further work examining both positive and negative welfare states with a view to the development of an automated system that can be used in precision livestock farming to improve animal welfare.

The rapid development of multicopters has led to many security problems. In order to prevent multicopters from invading restricted areas or famous buildings, an Anti-UAV Defense System (AUDS) has been developed and become a research topic of interest. Topics under research in relation to this include electromagnetic interference guns for unmanned aerial vehicles (UAVs), high-energy laser guns, US military net warheads, and AUDSs with net guns. However, these AUDSs use either manual aiming or expensive radar to track UAVs. This paper proposes a dual-axis rotary platform with UAV automatic tracking. The tracking platform uses visual image processing technology to track and lock the dynamic displacement of a UAV. When a target UAV is locked, the system uses a nine-axis attitude meter and laser rangefinders to measure its flight altitude and calculates its longitude and latitude coordinates through sphere coordinates to provide UAV monitoring for further defense or attack missions. Tracking tests of UAV flights in the air were carried out using a DJI MAVIC UAV at a height of 30 meters to 100 meters. It was set up for UAV image capture and visual recognition for tracking under various weather conditions by a thermal imager and a full-color camera respectively. When there was no cloud during the daytime, the images captured by the thermal imager and full-color camera provide a high-quality image recognition result. However, under dark weather, black clouds will emit radiant energy and seriously affect the capture of images by a thermal imager. When there is no cloud at night, the thermal imager performs well in UAV image capture. When the UAV is tracked and locked, the system can effectively obtain the flight altitude and longitude and latitude coordinate values.

Road conditions, often degraded by insufficient maintenance or adverse weather, significantly contribute to accidents, exacerbated by the limited human reaction time to sudden hazards like potholes. Early detection of distant potholes is crucial for timely corrective actions, such as reducing speed or avoiding obstacles, to mitigate vehicle damage and accidents. This paper introduces a novel approach that utilizes perspective transformation to enhance pothole detection at different distances, focusing particularly on distant potholes. Perspective transformation improves the visibility and clarity of potholes by virtually bringing them closer and enlarging their features, which is particularly beneficial given the fixed-size input requirement of object detection networks, typically smaller than the raw image resolutions captured by cameras. Our method automatically identifies the region of interest (ROI)—the road area—and calculates the corner points to generate a perspective transformation matrix. This matrix is applied to all images and corresponding bounding box labels, enhancing the representation of potholes in the dataset. This approach significantly boosts detection performance when used with YOLOv5-small, achieving a 45.7\% improvement in average precision (AP) at IoU thresholds of 0.5 to 0.95 for a single class, and notable improvements of 30.3\%, 78.6\%, and 278\% for near, medium, and far pothole classes, respectively, after categorizing them based on their distance. This work is the first to employ perspective transformation specifically for enhancing the detection of distant potholes.

To date, the development of unmanned technologies using autonomous underwater vehicles (AUV) has become an urgent demand for solving the problem of inspection of industrial sub-sea structures. A key issue here is precise localization of AUVs relative to underwater objects. However, the well-known challenges of recognizing and coordinating underwater objects (impossibility to use GPS, undercurrents, etc.) do not allow high-precision navigation if based solely on a standard suite of AUV navigation tools (sonars, etc.). An alternative is a technology of processing of optical images that, at short distances, can provide higher accuracy of AUV navigation compared to a technology of acoustic measurement processing. In this study, in the context of the problem of subsea production system inspection, we propose a method to recognize underwater objects and provide coordinate reference of AUV to them, based on stereo-image processing. Its distinctive features are the use of a non-standard technique for generating a geometric model of an object from its views (foreshortening) taken from positions of a pre-made overview trajectory, the use of various characteristic geometric elements when recognizing objects, and the original algorithms for comparing visual data of the inspection trajectory with an a priori model of the object.

Identifying punch types as part of punch kinematic analysis is a key component for both coaches and players, providing critical insights into the variety and effectiveness of punches, which are essential for refining overall performance. Existing research indicates that punch type classification typically relies on data from either wearables or video data. Both approaches have their merits and demerits but require a significant amount of labeled data for modeling the classification algorithm. In this work, we propose a novel approach that significantly reduces the labeling effort by one-sixth by leveraging video data to enhance the training information of sensor data. This enriched training phase enables more accurate punch classification with reduced labeled data. Subsequently, the real-time punch identification and classification use only the wearable data. On a real dataset, with 15% of the labeling efforts, our approach achieved an accuracy rate of 91.41% for rear hand punch recognition and 91.91% for lead hand punch recognition, along with 92.33% and 94.56% for punch classification, respectively. Through the development of our Smart Boxer system, we aim to revolutionize punch analytics in boxing, providing comprehensive insights while enhancing the overall training and competitive experience for athletes and enriching fan engagement with the sport.

In computer vision, human action recognition is a hot topic, popularized by the development of deep learning. Current models have achieved high accuracy results on public datasets. Despite this success, they require significant computational resources for training. Given that transfer learning based techniques allow reusing what other models have already learned and training models with less computational resources, in this work we propose using a transfer learning based approach for action recognition in videos. We describe a methodology for human action recognition using transfer learning techniques in a custom dataset. The proposed method consists of four stages: 1) human detection and tracking, 2) video preprocessing, 3) feature extraction (using pretrained models with ImageNet), and 4) action recognition using a two-stream model consisting of TCNs, LSTMs, and CNNs layers. The custom dataset is imbalanced with 189, 390, 490, 854, and 890 videos per class, respectively. For feature extraction, we analyzed the performance of seven pretrained models: Inception-v3, MobileNet-v2, MobileNet-v3-L, VGG-16, VGG-19, Xception, and ConvNeXt-L. We show that the best results were obtained with the last one. Finally, using pretrained models for feature extraction allowed training in a PC with a single GPU with an accuracy of 94.9%.

In recent years, Synthetic Aperture Radar (SAR) has received attention in the Automatic Target Recognition (ATR) field due to its ability to acquire images both day and night, combined with the superior resolution offered by the latest sensors. However, despite advances, there remains a deficiency in dataset quality and quantity. This has necessitated the use of simulated data to overcome challenges associated with obtaining large quantities of high-quality images of targets and their subsequent labeling. In this study, we trained a Convolutional Neural Network (CNN) using a simulated dataset, subsequently employing it to classify real images from the MSTAR dataset. Traditional training methodologies often struggle with generalization when confronted with domain disparities between simulated and real data. To mitigate this, we exploited Margin Disparity Discrepancy (MDD), a domain adaptation technique not previously explored in SAR context. Findings from this study showcased a 13% increase in classification accuracy and improved generalization using MDD. Through Explainable AI (XAI) techniques such as t-SNE and other metrics like Kullback Leibler Divergence (KLD), we revealed a more distinct alignment of feature spaces between the two domains after domain adaptation, emphasizing MDD’s role in enhancing model accuracy and generalization.

Medical image analysis is an important branch in the field of medicine, which mainly uses image processing and analysis techniques to interpret and diagnose medical image data. Medical image data helps doctors to effectively observe and diagnose patients' body structures, tissues and lesions. Medical image analysis has been an important research area in the medical field, and it is important for disease diagnosis, treatment planning, and condition monitoring. In recent years, the rapid development of deep learning and computer vision technologies has contributed greatly to the automation, multimodal data fusion, real-time application, and accuracy improvement of medical image analysis. In addition, the development of deep learning has given rise to some new research areas in medical image analysis, such as Generative Adversarial Networks (GANs) for synthetic medical images, self-supervised learning for unsupervised feature learning, and neural network interpretability. In this paper, we will introduce some optimisation methods for medical images which are effective in improving the accuracy, efficiency and reliability of medical image analysis.

The use of visual signals in horticulture has attracted significant attention and encompassed a wide range of data types such as 2D images, videos, hyperspectral images, and 3D point clouds. These visual signals have proven to be valuable in developing cutting-edge computer vision systems for various applications in horticulture, enabling plant growth monitoring, pest and disease detection, quality and yield estimation, and automated harvesting. However, unlike other sectors, developing deep learning computer vision systems for horticulture encounters unique challenges due to the limited availability of high-quality training and evaluation datasets necessary for deep learning models. This paper investigates the current status of vision systems and available data in order to identify the high-quality data requirements specific to horticultural applications. We analyse the impact of the quality of visual signals on the information content and features that can be extracted from these signals. To address the identified data quality requirements, we explore the usage of a deep learning-based super-resolution model for generative quality enhancement of visual signals. Furthermore, we discuss how these can be applied to meet the growing requirements around data quality for learning-based vision systems. We also present a detailed analysis of the competitive quality generated by the proposed solution compared to cost-intensive hardware-based alternatives. This work aims to guide the development of efficient computer vision models in horticulture by overcoming existing data challenges and paving a pathway forward for contemporary data acquisition.

Research has been growing on object detection using semi-supervised methods in past few years. We examine the intersection of these two areas for floor-plan objects to promote the research objective of detecting more accurate objects with less labelled data. The floor-plan objects include different furniture items with multiple types of the same class, and this high inter-class similarity impacts the performance of prior methods. In this paper, we present Mask R-CNN based semi-supervised approach that provides pixel-to-pixel alignment to generate individual annotation masks for each class to mine the inter-class similarity. The semi-supervised approach has a student-teacher network that pulls information from the teacher network and feeds it to the student network. The teacher network uses unlabeled data to form pseudo-boxes, and the student network uses both unlabeled data with the pseudo boxes and labelled data as ground truth for training. It learns representations of furniture items by combining labelled and unlabeled data. On the Mask R-CNN detector with ResNet-101 backbone network, the proposed approach achieves mAP of 98.8%, 99.7%, and 99.8% with only 1%, 5% and 10% labelled data, respectively. Our experiment affirms the efficiency of the proposed approach as it outperforms the fully supervised counterpart using only 10% of the labels.

Manual visual inspections typically conducted after an earthquake are high-risk, subjective, and time-consuming. Delays from inspections often exacerbate the social and economic impact of the disaster on affected communities. Rapid and autonomous inspection using images acquired from unmanned aerial vehicles offer the potential to reduce such delays. Indeed, a vast amount of re-search has been conducted toward developing automated vision-based methods to assess the health of infrastructure at the component and structure level. Most proposed methods typically rely on images of the damaged structure, but seldom consider how the images were acquired. To achieve autonomous inspections, methods must be evaluated in a comprehensive end-to-end manner, incorporating both data acquisition and data processing. In this paper, we leverage recent advances in computer generated imagery (CGI) to construct a 3D synthetic environment for simulation of post-earthquake inspections that allows for comprehensive evaluation and valida-tion of autonomous inspection strategies. A critical issue is how to simulate and subsequently render the damage in the structure after an earthquake. To this end, a high-fidelity nonlinear finite element model is incorporated in the synthetic environment to provide a representation of earthquake-induced damage; this finite element model, combined with photo-realistic rendering of the damage, is termed herein a physics-based graphics models (PBGM). The 3D synthetic en-vironment with PBGMs provide a comprehensive end-to-end approach for development and validation of autonomous post-earthquake strategies using UAVs, including: (i) simulation of path planning of virtual UAVs and image capture under different environmental conditions; (ii) au-tomatic labeling of captured images, potentially providing an infinite amount of data for training deep neural networks; (iii) availability of the ground truth damage state from the results of the finite-element simulation; and (iv) direct comparison of different approaches to autonomous as-sessments. Moreover, the synthetic data generated has the potential to be used to augment field datasets. To demonstrate the efficacy of PBGMs, models of reinforced concrete moment-frame buildings with masonry infill walls are examined. The 3D synthetic environment employing PBGMs is shown to provide an effective testbed for development and validation of autonomous vision-based post-earthquake inspections that can serve as an important building block for ad-vancing autonomous data to decision frameworks.

The relationship between the street environment and the health, education, mobility, and criminal behaviors of its citizens has long been investigated by economists, sociologists and urban planners. Home buyers were found to pay a premium for better street appearance. Prior studies considering streetscapes mainly focus on objective measures such as the number of nearby trees, the tree canopy area, or the view index of physical features such as greenery, sky or building. However, subjective perceptions may have complex or subtle relationships to physical features, individual physical features or simply summing them up do not capture people’s comprehensive perception. In contrast, this study proposed a new approach for the urban-scale application to quantify both subjectively and objectively measured streetscape scores for six important perception qualities, namely Greenness, Walkability, Safety, Imageability, Enclosure, and Complexity. Built on prior quantitative studies in urban design quality and emerging applications in deep learning and open source street view imagery for urban perceptions, we integrated existing frameworks to (1) effectively collect and evaluate both subjectively and objectively- measured perceptions; (2) investigate the coherence and divergence in ML-predicted subjective scores and formula-derived objective scores; and (3) compare their effects in affecting house prices taking Shanghai as a case study using a large-scale dataset on home transactions. The results implied: first, the percentage increase in sales price attributable to street scores is significant for both subjective and objective measurements. In general subjective scores explained more variance over structural attributes and objective scores in hedonic price model. Particularly, objective Greenness, subjective Safety and Imageability scores positively affected house prices. Second, for Greenness and Imageability scores, the subjective and objective measures exhibited opposite signs in affecting house prices, which implied that there might be mechanisms related to the psychological, social-demographical characteristics of street users that have not been fully incorporated by objective measures that taking view indices or recombination of them. In addition, certain objective measure might outperform subjective counterpart when the connotation of the perception is self-evident and not complicated, for example the Greenness. For those concepts were not familiar to the average person, subjective framework exhibits better performance. This is the first study comprehensively expanding hedonic price method with both subjectively and objectively measured streetscape qualities. It suggested that city authorities could levy a street environment tax to compensate the public budget invested in street environment where developers secured benefits from a price premium. This study enriches our understanding of the economic values of the subjective and objective measures street qualities. It sheds light on promising future study areas where the coherence and divergence of the two measurements should be further stressed.

As the Autonomous Vehicle (AV) industry is rapidly advancing, classification of non-motorized (vulnerable) road users (VRUs) becomes essential to ensure their safety and to smooth operation of road applications. The typical practice of non-motorized road users’ classification usually takes numerous training time and ignores the temporal evolution and behavior of the signal. In this research effort, we attempt to detect VRUs with high accuracy be proposing a novel framework that includes using Deep Transfer Learning, which saves training time and cost, to classify images constructed from Recurrence Quantification Analysis (RQA) that reflect the temporal dynamics and behavior of the signal. Recurrence Plots (RPs) were constructed from low-power smartphone sensors without using GPS data. The resulted RPs were used as inputs for different pre-trained Convolutional Neural Network (CNN) classifiers including constructing 227×227 images to be used for AlexNet and SqueezeNet; and constructing 224×224 images to be used for VGG16 and VGG19. Results show that the classification accuracy of Convolutional Neural Network Transfer Learning (CNN-TL) reaches 98.70%, 98.62%, 98.71%, and 98.71% for AlexNet, SqueezeNet, VGG16, and VGG19, respectively. The results of the proposed framework outperform other results in the literature (to the best of our knowledge) and show that using CNN-TL is promising for VRUs classification. Because of its relative straightforwardness, ability to be generalized and transferred, and potential high accuracy, we anticipate that this framework might be able to solve various problems related to signal classification.

In order to design efficient and sustainable office spaces and to automate lighting, heating and air circulation in these facilities, solving the challenge of occupancy prediction is crucial. In office spaces where large areas need to be observed, multiple sensors must be used for full coverage. In these cases, it is normally important to keep the costs low, but also to make sure that the privacy of the people who use such environments are preserved. Low-cost and low-resolution heat (thermal) sensors can be very useful to build solutions that address these concerns. However, they are extremely sensitive to noise artifacts which might be caused by heat prints of the people who left the space or by other objects which are either using electricity or exposed to sunlight. There are some earlier solutions for occupancy prediction that employ low-resolution heat sensors, however, they have not addressed nor compensate for such heat artifacts. Therefore, in this paper, we present a low-cost and low-energy consuming smart space implementation to predict the number of people in the environment based on whether their activity is static or dynamic in time. We use a low-resolution (8×8) and non-intrusive heat sensor to collect data from an actual meeting room. We propose two novel workflows to predict the occupancy; one based on computer vision and one based on machine learning. Besides comparing the advantages and disadvantages of these different workflows, we use several state-of-the-art explainability methods in order to provide a detailed analysis of the algorithm parameters and how the image properties influence the resulting performance. Furthermore, we analyze noise resources which affect the heat sensor data. We hope that our analysis brings light into understanding how to handle very low-resolution heat images in these environments. The presented workflows could be used in various domains and applications other than smart offices, where occupancy prediction is essential, e.g., for elderly care.

Research on Traffic Light Recognition (TLR) has grown in recent years, primarily driven by the growing interest in autonomous vehicles development. Machine Learning algorithms have been widely used to that purpose. Mainstream approaches, however, require large amount of data to properly work, and as a consequence, a lot of computational resources. In this paper we propose the use of Expert Instruction (IE) as a mechanism to reduce the amount of data required to provide accurate ML models for TLR. Given an image of the exterior scene taken from the inside of the vehicle, we stand the hypothesis that the picture of a traffic light is more likely to appear in the central and upper regions of the image. Frequency Maps of traffic light location were thus constructed to confirm this hypothesis. The frequency maps are the result of a manual effort of human experts in annotating each image with the coordinates of the region where the traffic light appears. Results show that EI increased the accuracy obtained by the classification algorithm in two different image datasets by at least 15%. Evaluation rates achieved by the inclusion of EI were also higher in further experiments, including traffic light detection followed by classification by the trained algorithm. The inclusion of EI in the PCANet achieved a precision of 83% and recall of 73% against 75.3% and 51.1%, respectively, of its counterpart. We finally presents a prototype of a TLR Device with that expert model embedded to assist drivers. The TLR uses a smartphone as a camera and processing unit. To show the feasibility of the apparatus, a dataset was obtained in real time usage and tested in an Adaptive Background Suppression Filter (AdaBSF) and Support Vector Machines (SVMs) algorithm to detect and recognize traffic lights. Results show precision of 100% and recall of 65%.

The integration of sign language recognition systems into digital platforms has the potential to bridge communication gaps between the deaf community and the broader population. This paper introduces an advanced Israeli Sign Language (ISL) recognition system designed to interpret dynamic motion gestures, addressing a critical need for more sophisticated and fluid communication tools. Unlike conventional systems that focus solely on static signs, our approach incorporates both Deep Learning and Computer Vision techniques to analyze and translate dynamic gestures captured in real-time video. We provide a comprehensive account of our preprocessing pipeline, detailing every stage from video collection to the extraction of landmarks using MediaPipe, including the mathematical equations used for preprocessing these landmarks and the final recognition process. The dataset utilized for training our model is unique in its comprehensiveness and is publicly accessible, enhancing the reproducibility and expansion of future research. The deployment of our model on a publicly accessible website allows users to engage with ISL interactively, facilitating both learning and practice. We discuss the development process, the challenges overcome, and the anticipated societal impact of our system in promoting greater inclusivity and understanding.

The state-of-the-art smart city has been calling for an economic but efficient energy management over large-scale network, especially for the electric power system. It is a critical issue to monitor, analyze and control electric loads of all users in system. In this paper, we employ the popular computer vision techniques of AI to design a non-intrusive load monitoring method for smart electric energy management. First of all, we utilize both signal transforms (including wavelet transform and discrete Fourier transform) and Gramian Angular Field (GAF) methods to map one-dimensional current signals onto two-dimensional color feature images. Second, we propose to recognize all electric loads from color feature images using a deep neural network with multi-scale feature extraction and attention mechanism. Third, we design our method as a cloud-based, non-intrusive monitoring of all users, thereby saving energy cost during electric power system control. Experimental results on both public and our private datasets have demonstrated our method achieves superior performances than its peers, and thus supports efficient energy management over large-scale Internet of Things.

Cycle times at workstations in offsite construction factories fluctuate widely due to various influencing factors. Consequently, relying on average rates, such as length per unit of time, for estimating cycle times proves to be inaccurate, often leading to significant deviations between production schedules and actual operations. To address this issue, this study proposes an estimation system that leverages machine-learning-based prediction, statistical methods, 3D simulation, and computer vision to predict cycle times at the workstation level. Testing of the system on a semi-automated wood-wall framing workstation in a panelized construction factory shows that it reduces the mean absolute error and sum of errors by approximately 36% and 68%, respectively, compared to the fixed rate method. The results also highlight the efficacy of using computer vision data for training machine-learning models for cycle time estimation, the importance of identifying and understanding the factors influencing cycle times, and the impact of random delays on the accuracy of cycle time estimation systems.

Indoor localization (IL) is a significant topic of study with several practical applications. The area of IL has evolved greatly in recent years due to the introduction of numerous technologies such as WiFi, Bluetooth, cameras, and other sensors. Despite the growing interest in this field, there are numerous challenges and drawbacks that must be addressed to develop more accurate and sustainable systems for IL and its real-life applications. This review study gives an in-depth look into IL, covering the most promising artificial intelligence-based and hybrid strategies that have shown excellent potential in overcoming some of the limitations of classic methods. In addition, the paper investigates the significance of high-quality datasets and evaluation metrics in the design and assessment of IL algorithms. Furthermore, this overview study emphasizes the crucial role that machine learning techniques, such as deep learning and transfer learning, play in the advancement of IL. A focus on the importance of IL and the various technologies, methods, and techniques that are being used to improve it. Finally, The survey highlights the need for continued research and development to create more accurate and scalable techniques that can be applied across a range of industries, such as evacuation-egress routes, hazard-crime detection, smart occupancy-driven energy reduction and asset tracking and management.

Intertidal seagrass plays a vital role in estimating the overall health and dynamics of coastal environments due to its interaction with tidal changes. However, most seagrass habitats around the globe have been in steady decline due to human impacts, disturbing the already delicate balance in environmental conditions that sustain seagrass. Miniaturization of multi-spectral sensors has facilitated very high resolution mapping of seagrass meadows, which significantly improve the potential for ecologists to monitor changes. In this study, two analytical approaches used for classifying intertidal seagrass habitats are compared: Object-based Image Analysis (OBIA) and Fully Convolutional Neural Networks (FCNNs). Both methods produce pixel-wise classifications in order to create segmented maps, however FCNNs are an emerging set of algorithms within Deep Learning with sparse application towards seagrass mapping. Conversely, OBIA has been a prominent solution within this field, with many studies leveraging in-situ data and multiscale segmentation to create habitat maps. This work demonstrates the utility of FCNNs in a semi-supervised setting to map seagrass and other coastal features from an optical drone survey conducted at Budle Bay, Northumberland, England. Semi-supervision is also an emerging field within Deep Learning that has practical benefits of achieving state of the art results using only subsets of labelled data. This is especially beneficial for remote sensing applications where in-situ data is an expensive commodity. For our results, we show that FCNNs have comparable performance with standard OBIA method used by ecologists, while also noting an increase in performance for mapping ecological features that are sparsely labelled across the study site.

Purpose The gradual increase in geriatric issues and global imbalance of the ratio between patients and healthcare professionals has created a demand for intelligent systems with the least error-prone diagnosis results to be used by less medically trained persons and save clinical time. This paper aims at investigating the development of an image-based colourimetric analysis. The purpose of recognising such tests is to support wider users to begin a colourimetric test to be used at homecare settings, telepathology, etc. Design/methodology/approach The concept of an automatic colourimetric assay detection is delivered by utilising two cases. Training Deep Learning (DL) models on thousands of images of these tests using transfer learning, this paper i) classifies the type of the assay, and ii) classifies the colourimetric results. Findings This paper demonstrated that the assay type can be recognised using DL techniques with 100% accuracy within a fraction of a second. Some of the advantages of the pre-trained model over the calibration-based approach are robustness, readiness and suitability to deploy for similar applications within a shorter period of time. Originality/value To the best of our knowledge, this is the first attempt to provide Colourimetric Assay Type Classification (CATC) using DL. Humans are capable to learn thousands of visual classifications in their life. Object recognition may be a trivial task for humans, due to photometric and geometric variabilities along with the high degree of intra-class variabilities it can be a challenging task for machines. However, transforming visual knowledge into machines, as proposed, can support non-experts to better manage their health and reduce some of the burdens on experts.

Autonomous aerial refueling technology is an effective solution to extend flight duration of unmanned aerial vehicles, and also a great challenge due to its high risk. For autonomous probe-and-drogue refueling tasks, relative navigation to provide relative position between the receiver aircraft and the refueling drogue is the first and essential step, and vision-based method is the most frequently used. A new monocular vision navigation sensor with image processing strategy consisting of the drogue detection method and the tracking method is developed for autonomous aerial refueling in this paper. In the drogue detection method, thresholding and mathematical morphology method are adopted to eliminate image interference, and contours extraction method is applied to obtain all contours, which are then subsequently checked to achieve target contour of drogue. In the tracking method, a rectangle of interest (ROI) of current frame image is determined by positioning results of last frame, and then processed by the previous drogue detection method. Finally, the proposed image processing strategy in monocular vision navigation sensor is validated using real flight images, which are captured from an autonomous aerial refueling testbed using a micro six-rotor aircraft as receiver aircraft.

Due to Motor Neurone Diseases, a large population remains disabled worldwide, negatively impacting their independence and quality of life. This typically involves a weakness in the hand and forearm muscles, making it difficult to perform fine motor tasks such as writing, buttoning a shirt, or gripping objects. This project presents a vision-enabled rehabilitation hand exoskeleton to assist disabled persons in their hand movements. The design goal was to create an accessible tool to help with a simple interface requiring no training. This prototype is built on a commercially available glove where a camera and embedded processor were integrated to help open and close the hand, using air pressure, thus grabbing an object. An accelerometer is also implemented to detect the characteristic hand gesture to release the object when desired. This passive vision-based control differs from active EMG-based designs as it does not require individualized training. Continuing the research will reduce the cost, weight, and power consumption to facilitate mass implementation.

This research explores the transformative impact of artificial intelligence (AI) on personalizing marketing campaigns, highlighting how AI technologies are reshaping modern marketing practices. AI tools such as machine learning, natural language processing (NLP), predictive analytics, deep learning, and computer vision are integral to enhancing customer engagement and optimizing marketing strategies. Machine learning algorithms enable precise customer segmentation and personalized recommendations, leading to more targeted and effective marketing messages. NLP facilitates the analysis of customer sentiment and enhances communication through advanced language processing and AI-powered chatbots. Predictive analytics and deep learning provide businesses with the capability to forecast customer behaviors and recognize complex data patterns, further refining marketing efforts. Computer vision technology contributes to personalization through innovations like visual search and augmented reality, creating immersive customer experiences. Despite the numerous benefits, the adoption of AI in marketing presents several challenges. Data privacy and security concerns, integration complexities, and the potential for algorithmic bias are critical issues that need to be addressed. Ethical considerations, including transparency and fairness in data usage, are also paramount to maintaining customer trust and ensuring responsible AI application. The future of AI in marketing promises continued advancements, with trends such as hyper-personalization and enhanced integration with technologies like the Internet of Things and voice search expected to drive further innovation. Overall, AI significantly enhances marketing effectiveness by offering tailored solutions and strategic insights while necessitating a thoughtful approach to ethical and practical challenges.

The Eurasian grapevine (Vitis vinifera L.) is the most widely grown horticultural crop in the world and is important for the economy of many countries. In the wine production chain, grape varieties play an important role as they directly influence the authenticity and classification of the product. Identifying the different grape varieties is therefore fundamental for quality control and inspection activities, as well as for regulating production. Currently, ampelography and molecular analysis are the main approaches to identifying grape varieties. However, both methods have limitations. Ampelography is subjective and prone to errors and is experiencing enormous difficulties as ampelographers are increasingly scarce. On the other hand, molecular analyses are very demanding in terms of cost and time. In this scenario, Deep Learning (DL) and Machine Learning (ML) methods have emerged as a classification alternative to deal with the scarcity of ampelographs and avoid molecular analyses. In this study, the most recent and current methods for identifying grapevine varieties using DL classification-based approaches are presented through a systematic literature review. The classification pipeline of the 31 studies found in the literature was described, highlighting its pros and cons. Most of the studies used DL-based models trained with leaf images acquired in a controlled environment at a maximum distance of 1.2 metres to classify grape varieties. In addition, there is a large gap between practical applications and the datasets used: a great lack of varieties, limited data acquired in the field and a lack of tests on plants under adverse conditions. Potential directions for improving this area of research were also presented.

All types of sports are potential application scenarios for automatic and real-time visual object and event detection. In rink hockey, the popular roller quad skate variant of hockey team sports, it is of great interest to automatically track player’s movements and positions, player’s sticks and, also, making other judgments, such as being able to locate the ball. In this work, we introduce a real-time pipeline composed by an object detection model, created specifically for rink hockey games, followed by a knowledge-based event detection module. Even in the presence of occlusions and quick motions, our deep learning object detection model effectively identifies and tracks, in real-time, important visual elements such as: ball; players; sticks; referees; crowd; goalkeeper; and goal. Using a curated dataset composed by a collection of videos of rink hockey, comprising 2525 annotated frames, we trained and evaluated the algorithm performance and compare it to state of the art object detection techniques. Our object detection model, based on YOLOv7, presents a global accuracy of 80%, and presents a good performance in terms of accuracy and speed, according to our results, making it a good choice for rink hockey applications. In our initial tests, the event detection module successfully detected one important event type in rink hockey games, the occurrence of penalties.

This research introduces a pioneering solution to the challenges posed by gastrointestinal tract (GI) cancer in radiation therapy, focusing on the imperative task of precise organ segmentation for minimizing radiation-induced damage. GI imaging has historically used manual demarcation, which is laborious and uncomfortable for patients. We address this by introducing the ResECA-U-Net deep learning model, a novel combination of the UNet and ResNet34 architectures. Furthermore, we further augment its functionality by incorporating the Efficient Channel Attention (ECA-Net) methodology. By utilizing data from the UW-Madison Carbone Cancer Centre, we carefully investigate several image processing techniques designed to capture critical local characteristics. With its foundation in computer vision concepts, the ResECA-U-Net model is excellent at extracting fine details from GI images. Sophisticated metrics such as intersection over union (IoU) and the dice coefficient are used to evaluate performance. Our study's outcomes demonstrate the effectiveness of the suggested method, yielding an impressive 96.27% Dice coefficient and 91.48% IoU. These results highlight the significant contribution that our strategy has made to the advancement of cancer therapy. Beyond its scientific merits, this work has the potential to significantly enhance cancer patients' quality of life and provide better long-term outcomes. Our work is a significant step towards automating and optimizing the segmentation process, which can potentially change how GI cancer is treated completely.

People often make mistakes, so we try to automate every aspect of our lives. Sports is no exception. While just over a decade ago humans were analyzing games, today this is being done by artificial intelligence. Due to rapid development over the past decade, neural networks are now faster, more accurate, and in some areas even better than their human counterparts. In this paper, we present an algorithm that can detect player statistics during an NBA broadcast. It also helps users better understand the game and the use of augmented reality. The algorithm detects players on the court, tracks their movements, and assigns them to their respective teams. Using homography estimation, we transform the players’ positions from a three-dimensional space in the video to a two-dimensional space on the playing field plane. We define a new algorithm that predicts the players’ actions and their statistics. The results show that the proposed method can effectively identify the players, their respective teams, and their positions. It can also analyze their actions with high accuracy.