Search Results (9,846)

Background: Integrating human neural signals with computational vision systems offers a promising route toward more robust visual recognition, yet supporting mixed-granularity recognition, where both coarse- and fine-grained categories must be distinguished within a unified system, remains challenging due to the heterogeneous feature spaces of electroencephalography (EEG) and visual data. Methods: We propose “Align and Fuse,” a two-stage Transformer-based framework. Stage 1 constructs a shared semantic space using a hardness-aware multimodal supervised contrastive loss with Hard Negative Weighting to explicitly target confusable class pairs. Stage 2 employs a multimodal Transformer with co-attention to fuse the aligned features for classification. Results: On the 80-class EEG-ImageNet benchmark, our framework achieved 91.12% Top-1 accuracy under a temporally separated control protocol, improving over the corresponding vision-only (89.08%) and Standard Transformer (89.95%) baselines. Under the original stratified random split, it achieved 92.56% Top-1 accuracy; on the 40-class EEGCVPR dataset, accuracy reaches 95.82%. Cross-subject experiments yield 90.92% average Top-1 accuracy on four unseen subjects, and Grad-CAM analysis suggests that aligned EEG signals shift the model’s attention toward semantically relevant regions. Conclusions: Coupling hardness-aware alignment with decoupled multimodal fusion supports EEG-augmented recognition by leveraging complementary stimulus-related information under the evaluated protocols. Because EEG features are required at inference time, the framework is positioned as a human-in-the-loop EEG-augmented recognition system rather than a standalone vision model. Full article

Autonomous driving systems rely on a sophisticated pipeline of artificial intelligence models to perceive, predict, and plan in dynamic environments. This review presents a systematic analysis of the machine learning and deep learning models underpinning vehicle autonomy, spanning classical convolutional neural networks (CNNs) for object detection and semantic segmentation to recurrent and Transformer-based architectures for trajectory prediction and motion planning. It also provides a critical examination of the autonomous vehicle sensor stack, including cameras, LiDAR, radar, ultrasonics, and GNSS/IMU as data acquisition systems, highlighting modality-specific AI challenges such as monocular depth estimation, 3D point cloud processing, and radar Doppler interpretation. The evolution of perception and decision-making pipelines is reviewed, contrasting modular architectures with end-to-end learning paradigms that directly map raw sensor data to control commands, and discussing their trade-offs in interpretability, safety assurance, and robustness to rare edge cases. We further survey specialized hardware accelerators and heterogeneous automotive SoCs designed to meet stringent real-time and power constraints. Industrial strategies are compared, including multi-modal sensor fusion and vision-centric approaches based on large-scale imitation learning. Finally, we identify open challenges related to robustness under adverse conditions, domain shift, causal ambiguity, and the need for interpretable and certifiable AI in safety-critical autonomous driving systems. Full article

Traditional manual grading of fresh chili peppers suffers from inconsistent quality control and low efficiency. To meet the demand for accurate fruit shape recognition during the post-harvest stage, this study proposes an intelligent recognition method based on an improved DenseNet-121 network. This approach facilitates the application of machine vision in agricultural sorting equipment. DenseNet-121 serves as the backbone network. The Convolutional Block Attention Module (CBAM) is introduced to enhance feature focus on fruit shapes. A regularization strategy (Dropout = 0.3, weight decay = 1 × 10⁻⁴) and a cross-entropy loss function with label smoothing (LS = 0.1) are integrated to optimize decision boundaries. These configurations prevent the model from overfitting to hard training labels and yield a robust classification architecture. Experimental results demonstrate that the proposed model achieves a precision of 90.09%, a recall of 89.60%, an F1-score (the harmonic mean of precision and recall) of 89.53%, and an overall accuracy of 89.74%. The model contains 7.09 M parameters and requires a single-frame inference time of 7.35 ms. Comprehensive evaluations indicate that the proposed model achieves an optimal balance among environmental noise robustness, prediction accuracy, and computational efficiency. Consequently, by maintaining high fine-grained classification accuracy alongside a low memory footprint and rapid inference speed, the model demonstrates strong potential for real-time deployment on resource-constrained edge devices within actual agricultural optical sorting equipment. Full article

Monitoring the dynamic response of structures subjected to operational loads is a key component of structural health assessment, providing valuable information for safety evaluation and maintenance planning. In the last decade, video-based measurements have received growing attention for modal identification and damage detection applications, offering a promising alternative to traditional sensor-based approaches. Unlike conventional monitoring systems, which provide discrete measurements and often require extensive instrumentation, computer vision techniques enable dense, non-contact measurements while reducing installation costs and accessibility constraints. Moreover, Motion Magnification algorithms can be combined with computer vision-based identification techniques to amplify displacements within selected frequency ranges, facilitating the detection of low-amplitude structural vibrations. In this work, a semi-automated methodology for structural identification is presented and validated through two experimental applications involving vibrating systems monitored with commercial cameras. The proposed framework combines computer vision algorithms, Motion Magnification (MM), correlation analysis, and Principal Component Analysis (PCA), the latter being adopted as a noise-reduction and dimensionality-reduction tool to extract the most informative features from large sets of time-histories. In contrast to previous studies primarily focused on damage detection and frequency evolution tracking, the present work specifically investigates the influence of key user-defined parameters on the reliability of the identified frequencies and provides practical calibration guidelines for future applications. The methodology was validated against reference measurements obtained from an optical monitoring system and it successfully identified the natural frequencies of the analysed structures with errors ranging from 0.84% to 1.75%. Sensitivity analyses performed on the region of interest size and position, as well as on the correlation threshold, demonstrated the robustness of the proposed workflow. The results confirm that the proposed approach represents a reliable, low-cost, and minimally invasive alternative to conventional dynamic monitoring techniques, while providing practical recommendations for its implementation in real-world structural health monitoring applications. Full article

Unmanned Aerial Vehicle (UAV)-based small object detection is a challenging computer vision task. It is constrained by two primary factors: UAV platforms have limited onboard computational resources, and high-altitude objects often have weak features that are easily overwhelmed by complex backgrounds. To address these challenges, we propose LUIM-YOLO. First, a Lightweight Multi-Scale Feature Enhancement (LMSFE) module integrates parallel multi-scale convolutions with attention to strengthen small and low-contrast object feature extraction. Second, an Adaptive Multi-Scale Bottleneck (AMSB) module enhances key semantic features of small objects and spatial correlation of medium-scale objects. Third, an Enhanced Cross-layer Compensation Feature Pyramid Network (ECC-FPN) constructs cross-level interaction pathways to improve small object position and scale perception. Experimental results on VisDrone2019 show that compared with YOLOv8n, LUIM-YOLO reduces parameters by 57% and improves mAP@50 by 12.9%. Additional full-validation-set PyTorch inference tests on NVIDIA Jetson Orin show that LUIM-YOLO achieves 88.19 ms/image in FP32, indicating a parameter-efficient accuracy-oriented design with edge deployment potential. Full article

Background/Objectives: Manual Pap smear analysis for cervical cancer screening is limited by inter-observer variability, time constraints, and restricted expert availability. Although convolutional neural networks (CNNs) have automated cervical cell classification, they remain limited in modeling long-range spatial dependencies and often lack clinical interpretability. Methods: In this study, Vision Transformer (ViT) architectures were systematically optimized to enhance automated cervical cancer screening and improve interpretability. The Herlev dataset (917 images: 242 normal, 675 abnormal) was utilized to optimize ViT-Tiny, a lightweight ViT architecture designed for reduced computational complexity, through a comprehensive evaluation of augmentation strategies, class weighting, and hyperparameters. Results: The optimal configuration achieved a cross-validation accuracy of approximately 95% (94.89% for the best replicated configuration), in which random horizontal flipping and class weighting (0.7 × 1.3) were identified as most effective. Gradient-weighted Class Activation Mapping (Grad-CAM) analysis confirmed that model attention corresponded to clinically relevant morphological features, including nuclei regions, cell boundaries, and chromatin texture, which align with cytopathological criteria. Conclusions: These findings indicate that Vision Transformers can deliver accurate and interpretable decision support for cervical cancer screening by combining competitive classification performance with attention-based transparency relevant to medical AI. Further validation on larger, multi-center datasets remains necessary before clinical deployment. Full article

Image segmentation is a critical step in computer vision systems for phytosanitary diagnosis in cacao production. However, the reliability of classical segmentation methods remains insufficiently assessed under real field conditions, where images captured under non-standardized conditions are often affected by variable illumination, complex backgrounds, partial occlusions, and chromatic similarity between cacao pods and surrounding vegetation. This study compares global thresholding, K-means clustering, and GrabCut using 343 cocoa pod images captured in uncontrolled agricultural environments with non-standardized mobile devices; low-resolution images were retained to preserve external validity. Robustness was assessed on the full dataset using unsupervised structural metrics, including the segmented area ratio (AS), the largest component ratio (LCR), and the catastrophic failure rate (FC), while accuracy was validated on 50 manually annotated images using Intersection over Union (IoU). Wilcoxon signed-rank tests indicated statistically significant differences among methods. GrabCut achieved the best performance (IoU = 0.814), high structural coherence (LCR = 0.985), and a low catastrophic failure rate (FC = 1.7%). In contrast, K-means showed severe fragmentation and instability, whereas global thresholding was highly sensitive to illumination variability and complex backgrounds. These results indicate that GrabCut provides a robust training-free baseline for cocoa pod segmentation under uncontrolled field conditions, particularly for offline phytosanitary analysis where annotated datasets, supervised training, or GPU-based deployment are limited. Full article

Background/Objectives: Impacted third molars may adversely affect adjacent second molars, leading to pathological conditions such as external root resorption and dental caries. Accurate assessment of these interactions is important for treatment planning and clinical decision-making. Although cone-beam computed tomography (CBCT) provides detailed three-dimensional imaging, image interpretation remains challenging. Recent advances in artificial intelligence have enabled automated radiographic analysis using deep learning methods. Methods: This retrospective study included 162 CBCT scans obtained from patients aged 18–75 years. A total of 306 third molar–second molar units were evaluated. Based on radiographic findings, interactions were categorized as independent, contact, or resorption. Several deep learning architectures were developed and evaluated, including conventional convolutional neural networks (CNNs), attention-based CNNs, and Vision Transformer (ViT) models. Performance was assessed using standard classification metrics, and an ensemble approach was applied to improve predictive stability. Results: Attention-based and Transformer-based models generally outperformed conventional CNN architectures. These models achieved better discrimination among the defined classes and demonstrated superior overall performance. The ensemble model produced the most reliable results, achieving the highest macro-area under the curve (macro-AUC) values. Distinguishing contact cases from independent cases was the most challenging task, whereas resorption cases were identified more consistently across different models. Conclusions: Transformer-based deep learning models showed promising performance for CBCT-based assessment of third molar–second molar interactions. Ensemble learning further improved classification reliability and robustness. These findings suggest that artificial intelligence-assisted systems may support early detection of third molar-related pathological changes and contribute to more accurate radiological evaluation and clinical decision-making. Full article

The construction industry continues to face high levels of accidents despite the use of various safety training approaches, highlighting the need for more effective and responsive methods. This study examines the role of Generative Artificial Intelligence (GenAI) in potentially improving construction safety training by exploring the development of training practices and identifying the shortcomings of existing approaches. A systematic literature review (SLR) was conducted to analyse safety training methods and emerging GenAI applications, followed by validation interviews with industry experts in South Australia to ensure practical relevance. Emergent findings show that safety training has progressed through three main stages: instructor-led, digital and GenAI-enabled. However, instructor-led and digital approaches remain limited by non-interactive learning, limited flexibility to different learner needs, lack of real-time feedback and weak alignment with actual site conditions. In contrast, GenAI offers opportunities to support more interactive, personalised and context-aware training through technologies such as large language models (LLMs), adaptive learning systems, computer vision and scenario generation. Despite these benefits, significant challenges related to data quality, system reliability, ethical concerns and organisational readiness continue to affect implementation. Based on these findings, the study develops an integrated framework that links training evolution, key challenges and GenAI capabilities, providing practical guidance to improve safety training in construction. Full article

Accurate tomato detection in greenhouse imagery is essential for robotic harvesting, yield estimation, and crop monitoring, yet visual clutter, fruit overlap, partial occlusion, and variable illumination remain challenging for object detectors. Although attention modules are frequently used in agricultural vision studies to improve feature discrimination, their practical contribution is often reported without controlled comparison against strong baseline detectors. This study presents a reproducible and deployment-aware benchmark for single-class greenhouse tomato detection using 895 images with 4930 annotated tomato instances in PASCAL VOC format. The first experimental block used a fixed 70/20/10 split to compare Faster R-CNN, four attention-augmented Faster R-CNN variants, Cascade R-CNN with ResNet101-DCN-FPN, and YOLOv11s attention variants. A second extended protocol converted the annotations to YOLO format and evaluated YOLO-family detectors and RT-DETR-l under a stratified 70/15/15 split, including ablation, robustness, seed-stability, and deployment analyses. The annotation audit confirmed valid bounding boxes, no empty images, and a high proportion of small tomato instances. In the first block, attention integration did not consistently improve detection performance, whereas Cascade R-CNN achieved the highest accuracy with 92.80% ${\mathrm{mAP}}_{0.5}$ and 90.80% $F_1$-score. In the extended protocol, RT-DETR-l obtained the highest test accuracy with 91.49% ${\mathrm{mAP}}_{0.5}$ and 58.59% ${\mathrm{mAP}}_{0.5:0.95}$, while Final-YOLO11s achieved comparable performance with lower latency, reaching 91.42% ${\mathrm{mAP}}_{0.5}$, 58.37% ${\mathrm{mAP}}_{0.5:0.95}$, and 86.19% $F_1$-score. Across three seeds, Final-YOLO11s obtained a stable mean ${\mathrm{mAP}}_{0.5}$ of 90.84%. Robustness analysis showed that motion blur and Gaussian noise caused the largest degradation, whereas compact YOLO models exported reliably to ONNX and TensorRT. Overall, the results indicate that localization quality, robustness, latency, model size, stability, and export capability should be considered together, and that adding attention modules by default is less reliable than evidence-driven detector selection. Full article

Three-dimensional surface reconstruction is essential for accurately acquiring the external quality parameters of watermelons, such as size, volume, and defect area. Binocular stereo vision provides a low-cost and easily deployable solution for the single-view 3D reconstruction of watermelons. However, watermelons present highly similar surface textures, and as typical spheroid-like objects, the excessive angle between surface normals of edge regions and the camera optical axis leads to insufficient feature representation. Consequently, directly applying existing stereo matching algorithms often introduces matching ambiguities, and lightweight networks struggle to balance real-time performance with matching accuracy. This study focuses on the high-precision single-view point cloud generation of Kirin watermelons. To address these issues, we first construct a cross-modal, high-precision Kirin watermelon stereo matching dataset. Building upon the Fast-ACVNet+ architecture, we then propose MI-ACVNet, a lightweight stereo matching network tailored for high-precision watermelon point cloud acquisition. In the feature extraction stage, a Multi-Scale Stereo Feature Extraction (MSFE) module is adapted. By incorporating the re-parameterized network MobileOne and Epipolar-Enhanced Coordinate Attention (E2CA), MSFE improves the discriminative capability for weak and similar textures without compromising inference speed. For cost computation, a Coarse-to-Fine Cascaded Residual Correction (C2F-CRC) strategy is incorporated to construct a fine-grained cost volume via sub-pixel interpolation, enhancing the network’s ability to capture subtle surface fluctuations. Furthermore, a Semantics-Guided Region-Aware Loss (SGRA-Loss) is formulated, leveraging semantic masks to apply differentiated supervision weights across edge, center, and background regions to significantly improve edge matching accuracy. Ablation studies validate the effectiveness of the MSFE, C2F-CRC, and SGRA-Loss components. Compared to the baseline model, the full MI-ACVNet reduces the End-Point Error (EPE) by 19.5% and the Bad-0.5 error rate by 34.5% in the watermelon region. Furthermore, when compared against five mainstream algorithms (StereoNet, AANet, HSMNet, LightStereo-L, and NMRF-swint), MI-ACVNet achieves state-of-the-art performance: EPE and Bad-0.5 are reduced to 0.091 pixels and 1.159%, respectively, with a single-frame inference time of only 46 ms. The average depth error of the reconstructed point clouds is merely 0.26 mm. By ensuring both real-time efficiency and high-precision depth estimation, this method demonstrates promising potential for deployment in industrial Kirin watermelon sorting lines, driving sorting equipment toward higher precision and intelligence. Full article

To address the challenges of severe optical attenuation and dynamic feature extraction for moving target depth retrieval in complex underwater remote sensing environments, this paper proposes a dynamic target depth estimation method based on multi-source data fusion. Taking optical RGB imagery and neuromorphic vision (NeuroIV) data as joint inputs, the proposed method constructs a three-channel feature extraction and fusion network. By leveraging a hypergraph structure, it establishes association weights between dynamic (temporal) and static (spatial) nodes to capture spatiotemporal correlations. To efficiently process the high-dimensional multi-modal data, the traditional dot-product attention is replaced with element-wise multiplication, significantly reducing computational complexity. Furthermore, a lightweight deformable attention pyramid (DAP) and diffusion model is introduced to refine depth image edges, effectively suppressing discontinuities and abruptness in the estimation results. Compared to single-modality optical imagery, the fused multi-modal data yields a superior signal-to-noise ratio and foreground contrast, achieving an improvement of over 20% in the MAE index. These results validate the effectiveness and superiority of the proposed multi-modal fusion strategy for dynamic target observation and depth retrieval in aquatic environments. Full article

Public YOLO model releases typically provide high-level architectural descriptions and headline benchmark results but offer limited empirical attribution of performance to individual blocks under controlled training conditions. This paper presents a modular, block-level analysis of YOLO26’s object detection architecture, detailing the design, function, and contribution of each component. We systematically examine YOLO26’s convolutional modules, bottleneck-based refinement blocks, spatial pyramid pooling, and position-sensitive attention mechanisms. Each block is analyzed in terms of objective and internal flow. In parallel, we conduct targeted ablation studies to quantify the effect of key design choices on accuracy (mAP@0.50:0.95) and inference latency under a fixed seed-0, COCO-only, fully specified training and benchmarking protocol. Experiments use the MS COCOdataset with the standard train2017 split (≈118 k images) for training and the full val2017 split (5 k images) for evaluation. The result is a self-contained empirical architectural-attribution reference that supports interpretability, reproducibility, and evidence-based architectural decision-making for real-time detection models. Beyond isolated ablations, we further synthesize the best-performing design choices into combined YOLO26n configurations and compare them against the default baseline. The best combined configuration improves mAP@0.50:0.95 from 0.3933 to 0.3969, while introducing only a marginal latency increase from 0.99 ms to 1.00 ms under TensorRT FP16 benchmarking. This analysis identifies an improved accuracy–latency trade-off and provides an incremental architectural configuration contribution supported by controlled experiments. The study is, therefore, framed as a controlled empirical analysis and configuration-refinement study of YOLO26, rather than as the proposal of a new detector family or a claim of universal detector superiority. Full article

Ammonia nitrogen is one of the most common environmental stressors in aquaculture water environments, and its accumulation can induce physiological disturbance, altered ventilation regulation, and abnormal behavioral responses in fish. To achieve non-invasive quantitative characterization of rainbow trout responses to ammonia nitrogen stress, this study developed a computer-vision-based framework for the integrated analysis of locomotor behavior and ventilation activity. Rainbow trout were exposed to four ammonia nitrogen concentrations: 0, 15, 30, and 60 mg/L. A total of 16 rainbow trout were used in this study, with an average body length of 14.0 ± 1.0 cm and an average body weight of 38.65 ± 2.42 g. The fish were assigned to four experimental aquaria, with four fish maintained in one aquarium for each TAN treatment. Stereo videos for locomotor behavior analysis and monocular mouth-region videos for ventilation analysis were simultaneously collected, and the final 5 min of each recording was analyzed. YOLOv11n, multi-object tracking, and stereo vision were used to extract three-dimensional position sequences of rainbow trout and calculate the amount of exercise, average swimming speed, and spatial distribution. Meanwhile, optical-flow analysis was applied to quantify mouth opening–closing motion and estimate ventilation frequency. The results showed that with increasing ammonia nitrogen concentration, rainbow trout locomotor behavior tended to be suppressed, with average swimming speed showing the clearest decrease, whereas ventilation frequency continuously increased. Average swimming speed decreased from 3.83 cm/s in the 0 mg/L group to 1.03 cm/s in the 60 mg/L group, while ventilation frequency increased from 84.91 breaths/min to 133.43 breaths/min. Compared with locomotor indicators, ventilation frequency showed a more stable response to changes in ammonia nitrogen concentration. This study achieved the synchronous quantification of rainbow trout locomotor behavior and ventilation activity, revealing a differentiated response pattern characterized by enhanced ventilation and suppressed locomotor behavior under acute ammonia nitrogen stress. These findings provide a methodological reference for fish stress assessment and risk warning in aquaculture environments. Full article

Visible and infrared (IR) image fusion has become an important strategy for improving computer vision performance under low illumination, occlusion, and some poor-visibility conditions. By integrating complementary textural information from visible images with thermal or IR cues, VIR fusion can enhance object localization, detection robustness, and facial-expression recognition (FER). This review examines VIR fusion techniques and datasets for computer vision applications, with object detection (OD) considered as a relatively mature scene-level task and FER considered as an emerging human-centered application. It summarizes major multimodal datasets, compares early-fusion approaches, including sensor- and feature-level fusion, with late-fusion approaches, including score- and decision-level fusion, and discusses representative machine learning and deep learning methods. The review also evaluates commonly used performance metrics and identifies current limitations, including dataset imbalance, sensor misalignment, limited demographic diversity in facial-expression datasets, computational complexity, and weak real-time generalization. Finally, key application areas, including surveillance, healthcare, remote sensing, autonomous systems, and human–computer interaction, are discussed. This review highlights the need for better-aligned multimodal datasets, standardized evaluation protocols, lightweight fusion architectures, and robust models capable of operating in dynamic real-world environments. Full article