Computer Vision

Falls bring about significant risks to individuals’ well-being and independence, prompting widespread public health concerns. Swift detection and even predicting the risk of falls are crucial for implementing effective measures to alleviate the adverse consequences associated with such incidents. This study presents a new framework for identifying and forecasting fall risks. Our approach utilizes a novel transformer model trained on 2D poses extracted through an off-the-shelf pose extractor, incorporating transfer learning techniques. Initially, the transformer is trained on a large dataset containing 2D poses of general actions. Subsequently, we freeze the majority of its layers and fine-tune only the last few layers using relatively smaller datasets for fall detection and prediction tasks. Experimental results indicate that our proposed method outperforms traditional machine learning (e.g., SVM, Decision Tree, etc.) and deep learning approaches (e.g., LSTM, CNN, ST-GCN, PoseC3D, etc.) in both fall detection and prediction tasks across various datasets.

Objective:Computer vision-based up-to-date accurate damage classification and localization are of decisive importance for infrastructure monitoring, safety, and the serviceability of civil infrastructure. Current state-of-the-art deep learning (DL)-based damage detection models, however, often lack superior feature extraction capability in complex and noisy environments, limiting the development of accurate and reliable object distinction. Method: To this end, we present DenseSPH-YOLOv5, a real-time DL-based high-performance damage detection model where DenseNet blocks have been integrated with the backbone to improve in preserving and reusing critical feature information. Additionally, convolutional block attention modules (CBAM) have been implemented to improve attention performance mechanisms for strong and discriminating deep spatial feature extraction that results in superior detection under various challenging environments. Moreover, additional feature fusion layers and a Swin-Transformer Prediction Head (SPH) have been added leveraging advanced self-attention mechanism for more efficient detection of multiscale object sizes and simultaneously reducing the computational complexity. Results: Evaluating the model performance in large-scale Road Damage Dataset (RDD-2018), at a detection rate of 62.4 FPS, DenseSPH-YOLOv5 obtains a mean average precision (mAP) value of 85.25 %, F1-score of 81.18 %, and precision (P) value of 89.51 % outperforming current state-of-the-art models. Significance: The present research provides an effective and efficient damage localization model addressing the shortcoming of existing DL-based damage detection models by providing highly accurate localized bounding box prediction. Current work constitutes a step towards an accurate and robust automated damage detection system in real-time in-field applications.

Face recognition systems are increasingly used in biometric security for convenience and effectiveness. However, they remain vulnerable to spoofing attacks, where attackers use photos, videos, or masks to impersonate legitimate users. This research addresses these vulnerabilities by exploring the Vision Transformer (ViT) architecture, fine-tuned with the DINO framework utilizing CelebA-Spoof, CASIA SURF, and a proprietary dataset. The DINO framework facilitates self-supervised learning, enabling the model to learn distinguishing features from unlabeled data. We compared the performance of the proposed fine-tuned ViT model using the DINO framework against traditional models, including CNN Model EfficientNet b2, EfficientNet b2 (Noisy Student), and Mobile ViT on the face anti-spoofing task. Numerous tests on standard datasets show that the ViT model performs better than other models in terms of accuracy and resistance to different spoofing methods. Our model’s superior performance, particularly in APCER (1.6%), the most critical metric in this domain, underscores its improved ability to detect spoofing relative to other models. Additionally, we collected our own dataset from a biometric application to validate our findings further. This study highlights the superior performance of transformer-based architecture in identifying complex spoofing cues, leading to significant advancements in biometric security.

Vision Transformer (ViT) is an image recognition model that uses transformer architecture, which has a numerous advantage over Convolution Neural Networks (CNN). It offers improved accuracy, scalability, flexibility, global context, and transferability. ViT can handle images of different sizes and aspect ratios, making it more versatile than CNN. It can process an entire image at once, allowing it to capture global context information and long-range dependencies. Additionally, ViTs pre-training on huge amounts of image data can be transferred to other image recognition tasks, making it a useful tool for transfer learning. This paper describes the differences between ViT and CNN and how ViT splits images into patches for classification. The positional encoding of different features is done in ViT to avoid the requirement of filters. Proposed implementation obtained final accuracy of prediction 93% for top-1 accuracy.

Northern Bangladesh is home to the most lychee growing, which has a major economic impact. Lychee output and quality are reduced by various leaf and fruit diseases. Deep learning is used to construct a disease detection system for lychee diseases to detect them early and accurately. We correctly identified 6,000 photos of healthy and unhealthy lychee foliage and fruits. Six deep learning algorithms—VGG16, CustomCNN, MobileNet, InceptionV3, ResNet50, and Vision Transformer— classified the photos. Normalization and augmentation improved model resilience during data preparation. F1-score, recall, accuracy, and precision were utilized to train and evaluate the models. Vision Transformer (ViT) scores 99.91% in accuracy, recall, and F1-score, outperforming the other models. This achievement shows that ViT can better recognize complex lychee disease traits. The ViT algorithm for disease detection can help Bangladeshi farmers identify illnesses quickly and accurately. Reduced lychee loss and improved fruit quality and attractiveness should boost lychee sales in domestic and international markets.

In order to achieve real-time classification and detection of various chicken parts, this study introduces an optimized Swin-Transformer method for the classification and detection of multiple chicken parts. It initially leverages the Transformer’s self-attention structure to capture more comprehensive high-level visual semantic information from chicken part images. The image enhancement technique was applied to the image in the preprocessing stage to enhance the feature information of the image, and the migration learning method was used to train and optimize the Swin-Transformer model on the enhanced chicken parts dataset for classification and detection of chicken parts. Furthermore, this model was compared to four commonly used models in object target detection tasks: YOLOV3-Darknet53, YOLOV3-MobileNetv3, SSD-MobileNetv3, and SSD-VGG16. The results indicated that the Swin-Transformer model outperforms these models with a higher mAP value by 1.62%, 2.13%, 5.26%, and 4.48%, accompanied by a reduction in detection time by 16.18 ms, 5.08 ms, 9.38 ms, and 23.48 ms, respectively. The method of this study fulfills the production line requirements while exhibiting superior performance and greater robustness compared to existing conventional methods.

With the expansion of the installed capacity of power system and the scale of power grid, the operation of various electrical equipment requires the reliability of power system to be improved continuously. Transformer is one of the key equipments in the power distribution system, and its running state is directly related to the reliability of the whole power system. Condition-based maintenance technology is to diagnose the equipment by using maintenance technology when the equipment is charged. On-line detection and fault detection technology of transformer can improve the reliability of equipment, reduce the maintenance cost of equipment and improve economic benefits. In this paper, a transformer fault detection algorithm based on computer vision and deep learning technology is proposed, and the effectiveness and feasibility of transformer computer vision fault online detection technology are verified from two aspects: fault detection scheme design and transformer fault detection system. The simulation results show that with the increase of the number of experiments, the accuracy of this algorithm is stable at about 95%, and the real-time wavelength tends to be stable, which is 15.85% higher than the traditional algorithm on average. This method can effectively detect the transformer fault, and at the same time, it can detect the latent fault that leads to the drift of transformer winding parameters.

Computer vision syndrome (CVS) is a major illness derived as a result of long-term improper usage of digital devices. It causes ocular as well as extraocular symptoms that may lead to blindness and cancers in severe cases. Therefore, early detection and diagnosis are crucial for preventing vision loss. Pupillometry offers an objective assessment of CVS, as continuous light exposure leads to characteristic changes in pupil size. In this paper, a transformer model algorithm for the segmentation of pupil, “ViT-Pupil” is employed as a preliminary step to evaluate pupil changes. This proposed method uses a CASIA iris degradation database for training the model, which contains gray-scale eye images that are of different sizes, shapes, and occlusions due to eyelashes. The model resulted in a validation dice coefficient of 0.93. A cross-domain generalization technique is carried out on a customized dataset containing RGB images captured using a low-resolution mobile camera under different light intensity levels using the proposed model. The customized images undergo preprocessing as an initial step. The segmentation result for the customized dataset indicates the superior performance of ViT-Pupil for a noisy, unclear images, which shows the robustness of the proposed method. Thus, the proposed method can be utilized to track the pupillary changes by integrating with wearable devices as a preliminary step for the identification of CVS.

Malware authors aim to target Windows Portable Executable (PE) files because these files are the common format for software applications. A Static Malware Detection Technique Using the PE File characteristics and features. A comparative analysis is done using grayscaled image converting of PE headers by ResNet and TabTransformer models against Directly using the extracted PE features in their original tabular format. The direct feature-based TabTransformer approach is found to be better than image-based methods as shown by experiments on benchmark dataset such as the Ransomware PE Header dataset and the Windows Malware Detection dataset. The findings reveal how they can help enhance static malware detection by better capturing the complex interactions of the PE file features. In addition, no transformation of the data into an image is needed when using TabTransformer.

Computer Vision Syndrome (CVS) has emerged as a critical health concern in the digital era due to prolonged exposure to screens, resulting in eye strain, dryness, and blurred vision.Traditional diagnostic methods rely on clinical examination, which can be invasive, time-consuming, and inaccessible for frequent monitoring.This paper introduces Eyelume, a real-time CVS detection and monitoring system that leverages Vision Transformers (ViT) for accurate pupil segmentation and pupillometry analysis.Unlike Convolutional Neural Networks (CNNs), ViTs capture global dependencies within visual data, offering robustness against low-quality and noisy eye images.The system enables users to upload eye images via a web interface, validates input quality, and computes pupil size variations to identify abnormal responses linked to CVS.Experimental evaluations demonstrate a segmentation accuracy of 99.6%, proving Eyelume's potential as a non-invasive, accessible, and effective tool for early CVS detection and digital eye health monitoring [1][2].

Modern computer vision systems need improved detection methods because their fast development requires them. An exploration takes place to study how attention mechanisms together with transformer networks can boost object detection results. The proposed architecture unites visual attention modules with transformer encoders to apply dual-path attention which handles spatial and channel-wise information processing jointly. A training process used the COCO dataset along with its 118,000 images while testing occurred with 5,000 test images. Through our complete framework we reached 94.7% mAP results in the test dataset which outperformed conventional CNN-based methods by 12.3%. The system delivered exceptional performance when detecting objects hidden by other objects and when dealing with changes in illumination intensity (92.1%). The system operated at 45 frames per second through normal GPU processing. The proposed architectural solution connects traditional computer vision systems to modern transformers networks through design that achieves high practical object detection accuracy with efficient computational performance.

Early and accurate detection of skin cancer, particularly melanoma, remains a critical challenge in computer-aided diagnosis, motivating the development of reliable and interpretable machine learning solutions. This study presents a comparative algorithmic analysis of deep learning models for automated skin cancer detection using dermoscopic images. Specifically, convolutional neural networks (CNNs) and Vision Transformers (ViTs) are implemented within a unified framework, employing transfer learning and standardized preprocessing techniques on a benchmark dataset. The proposed methodology incorporates data augmentation and class imbalance handling strategies, while model performance is evaluated using clinically relevant metrics, including accuracy, precision, recall, F1-score, and area under the ROC curve. In addition, explainability techniques such as Grad-CAM and attention visualization are employed to enhance model interpretability, and decision threshold analysis is conducted to assess trade-offs between sensitivity and specificity in melanoma detection. Experimental results demonstrate that CNN-based architectures achieve robust performance in capturing local spatial features, while transformer-based models provide competitive results through global contextual representation. However, variations are observed in model calibration and false-negative rates, which are critical for clinical deployment. Overall, the findings highlight the importance of combining algorithmic performance with interpretability and threshold optimization to support reliable and clinically meaningful computer-aided diagnosis systems.

Inattention in e-learning environments has emerged as a critical factor affecting student engagement and academic performance, particularly in the wake of widespread online education adoption. This paper proposes a modular and scalable architectural blueprint for a web-based application that enables real-time detection of student inattention levels using computer vision techniques and deep learning models. The design incorporates a facial expression analysis pipeline based on the Dataset for Affective States in E-Environments (DAiSEE) and considers two temporal modeling approaches: a Bi-directional long short-term memory (Bi-LSTM) and a Transformer-based architecture, for future implementation. A dual-phase validation plan is outlined to assess both the technical feasibility and educational relevance of the application, including quantitative model evaluation and qualitative feedback from instructors and students in real classroom settings. This proposal aims to provide a foundation for data-driven instructional support by enabling automated, real-time monitoring of inattention levels in online classes.