Search Results (5,820)

In low-light, occluded, and cluttered environments, single-modality RGB detectors are prone to false positives and missed detections. While infrared (IR) imaging provides relatively stable target visibility under poor illumination, it lacks texture and color information and is susceptible to background thermal noise and imaging variations. To address these limitations, this paper proposes an RGB–IR object detection network, named MCF-YOLO, consisting of three core components. First, the Cross-Modal Hierarchical Fusion (CMHF) module performs stage-wise alignment and fusion on multi-scale features, jointly modeling RGB texture details and IR thermal responses to exploit the structural and semantic complementarity between the two modalities. Second, the Soft Attention Regularization based on Attention Prior (SAR-AP) module derives attention priors from IR features to impose soft constraints on cross-modal attention maps. This mechanism helps the network maintain attention on target-relevant regions, thereby suppressing attention drift caused by low-light noise and complex backgrounds. Third, the Small-Object-Sensitive Detection Head (SOS-Head) processes high-resolution features to strengthen the representation of small targets, improving detection capability in long-range and occluded scenarios. In evaluations on two RGB–IR benchmarks—M3FD and VEDAI—MCF-YOLO achieves improvements of 2.7% in mAP@0.5 and 1.1% in mAP@0.5:0.95 on M3FD, and 5.4% and 4.4%, respectively, on VEDAI. These results suggest that consistency-guided cross-modal fusion and high-resolution small-target modeling are beneficial for RGB–IR detection in low-visibility and cluttered scenes. Full article

Adverse imaging conditions such as fog, rain, and low light degrade the reliability of vision-based Personal Protective Equipment (PPE) detection systems on construction sites, yet most existing models are trained under clear-weather assumptions. This paper introduces a physics-based weather augmentation framework integrated with the YOLOv8n architecture to improve PPE detection robustness under adverse environmental conditions. The original Color Helmet and Vest (CHV) dataset was expanded from 1330 clear-weather images to 6650 images across five conditions using four physically grounded augmentation models: the Koschmieder atmospheric scattering model for fog, the Garg–Nayar streak model for rain, gamma-corrected attenuation with Poisson–Gaussian noise for low light, and a PSF-based glare model for bright sunlight. The weather-resistant model, a clear-weather baseline, and an augmented baseline were evaluated on the same 665-image weather-augmented test set. The weather-resistant model achieves 89.2% mAP50, a 5.7 percentage-point improvement over the clear-weather baseline (83.5%), with a nearly four-fold improvement in cross-condition stability (standard deviation 1.5% vs. 5.7%). Under matched training-data volume, the weather-resistant model still outperforms a conventionally augmented baseline across all five simulated conditions, indicating that these gains stem from physics-based modeling rather than larger training-data volume. The largest gain occurs under low light, where mAP50 improves from 73.4% to 87.9%. Gradient-weighted Class Activation Mapping (Grad-CAM) analysis confirms that the weather-resistant model directs more attention toward PPE regions across all conditions, with the largest improvement under low light (+10.0 percentage points). The lightweight design (3.0 M parameters) and quantitative and qualitative validation on 205 annotated real-world construction site images under normal and low-light conditions provide preliminary evidence of practical applicability. Full article

Background/Objectives: Primary thyroid lymphoma (PTL), including diffuse large B-cell lymphoma (DLBCL) and mucosa-associated lymphoid tissue (MALT) lymphoma, share substantial overlap in ultrasound appearance with Hashimoto’s thyroiditis (HT), making preoperative differentiation challenging. This study aims to develop and validate a deep learning model based on B-mode ultrasound (BMUS) and color Doppler ultrasound (CDUS) for image-level differentiation of DLBCL, MALT lymphoma, and HT. Methods: This retrospective single-center study included 1294 ultrasound images from 290 patients (313 lesions) who underwent preoperative ultrasound examination at West China Hospital between September 2002 and September 2024. All images from the same lesion were assigned to the same data partition, and the dataset was split at the lesion level into training and test sets at an 8:2 ratio. A Frequency-Adaptive WT-ResNet model incorporating wavelet transform convolution and a frequency-adaptive gating mechanism was developed. The primary analysis was performed at the image level. The performance of the model was compared with that of three ultrasound physicians with different levels of experience. Grad-CAM was used for visual interpretation. An exploratory external validation was performed using an independent dataset from Sun Yat-sen Memorial Hospital. Results: In the test set, the model achieved a macro-average AUC of 0.927 (95% CI: 0.889–0.960), with class-specific AUCs of 0.899 for DLBCL, 0.946 for MALT lymphoma, and 0.937 for HT. The macro-average balanced accuracy was 0.866, compared with 0.827 for that of the best-performing senior physician. The exploratory validation set yielded a macro-average AUC of 0.796 (95% CI: 0.686–0.888), with class-specific AUCs of 0.806 for DLBCL, 0.825 for HT, and 0.756 for MALT lymphoma. Grad-CAM showed that the model focused on lesion-internal echotexture and lesion-transition regions with class-dependent patterns. Conclusions: A deep learning model based on BMUS and CDUS showed promising performance for image-level differentiation of DLBCL, MALT lymphoma and HT in a single-center retrospective cohort. The model outperformed three ultrasound physicians and may serve as a potential decision-support tool. However, the exploratory external validation results should be interpreted as preliminary, and larger multicenter cohorts remain necessary to confirm model generalizability. Full article

Background: Tongue diagnosis is a fundamental diagnostic method in traditional medicine. Studies restricted to single ethnic groups may introduce bias and limit the clinical applicability of digital tongue diagnosis across diverse populations. Objectives: This study examined differences in tongue image features between Japanese and Brazilian (Caucasian ancestry) participants using digital tongue diagnosis technology and explored potential influencing factors. Methods: Tongue images were collected from 143 Japanese and 116 Brazilian participants attending traditional medicine clinics in Japan and Brazil. An independently developed tongue image analysis system (TIAS) was employed to extract shape, texture (gray level co-occurrence matrix), color (L*a*b color space), and deep-learning derived features (crack, prickle, tooth-mark, peel, greasy coating, stasis). Statistical analyses and machine learning models with SHAP explainability were used to compare features and identify key classification parameters. Results: Significant inter-group differences were observed in tongue shape, texture parameters (particularly at the root and tip), color parameters (especially middle-a-mean, middle-b-mean, tip-a-mean, and tip-b-mean), and deep features. The Japanese group showed a markedly higher prevalence of greasy coating (72.03% vs. 41.38%, p < 0.001) and stasis. Machine learning analysis revealed that the b value in the middle region of the tongue (middle-b-mean) contributed most strongly to the classification of greasy coating. Conclusions: The digital tongue image analysis system enables accurate and objective quantification of tongue features. Pronounced ethnic differences exist, particularly in the distribution of greasy coating. The middle-b-mean has the greatest impact on greasy coating classification. These findings underscore the importance of considering ethnic background when developing digital tongue diagnosis systems. Full article

Accurate and robust detection of leaf diseases is a key enabler for precision agriculture and large-scale crop health monitoring. Despite the strong generalization of modern one-stage detectors (e.g., YOLOv8), two domain-specific challenges remain: (i) weak or blurry lesion boundaries hinder precise localization, and (ii) low color contrast between diseased and healthy tissues forces models to rely on subtle texture patterns rather than salient shapes. To tackle these challenges, we reframe the core agricultural disease detection task as the identification of “asymmetric morphological anomalies” and propose a domain-tailored enhancement framework. First, we introduce an Edge Enhancement Module (EEM) that explicitly strengthens boundary-aware representations. Inspired by the natural symmetry of healthy leaves, our EEM is specifically designed to capture symmetry-breaking boundary discontinuities and localized asymmetric edges caused by disease lesions. Our method enhances edge and texture cues that are indicative of disease lesions, which often exhibit local asymmetries and boundary discontinuities. The EEM includes a Differential Normalized Pooling Block (DNPB) that highlights edge responses through discrepancies between max pooling and average pooling, which also models cross-group edge correlations. Second, the Lightweight Texture-Sensitive Feature Enhancement (LTSFE) mechanism amplifies texture-discriminative channels under low-contrast conditions by leveraging complementary global statistics and efficient channel mixing, all with negligible computational overhead. We evaluated our method on a self-constructed dataset of 106,434 images with 225,640 annotations covering diverse crops. Experiments show that the proposed method achieves state-of-the-art accuracy (81.54% mAP@0.5:0.95) while maintaining real-time inference (142 FPS), consistently outperforming strong baselines. Ablations confirm the effectiveness and complementarity of EEM and LTSFE, demonstrating that domain-specific architectural design, inspired by biological symmetry, can substantially improve agricultural vision systems. Full article

Background/Objectives: Although vascular mechanisms are increasingly implicated in the etiology of sudden sensorineural hearing loss (SSNHL), the inability to directly visualize the labyrinthine artery remains a diagnostic obstacle. Sharing embryological and physiological parallels with the inner ear, the eye represents an accessible surrogate organ capable of reflecting systemic microvascular status. This study aimed to evaluate the diagnostic value of ocular hemodynamic and structural parameters in patients with acute unilateral idiopathic SSNHL. Methods: This prospective, comparative, cross-sectional study enrolled 30 patients with acute unilateral idiopathic SSNHL and 25 age and sex matched healthy controls. Three groups were defined: the affected eye, the contralateral eye, and the control eye. Retrobulbar hemodynamics (PSV, EDV, RI, PI) were assessed by color Doppler imaging; peripapillary choroidal thickness, RNFL, GCC+, and macular thickness by swept-source OCT; and macular microvascular perfusion by OCT angiography. Results: End diastolic velocity in the posterior ciliary arteries was significantly reduced in both patient eye groups relative to controls (p < 0.001), while RI and PI were significantly elevated (p = 0.001 and p = 0.004, respectively). Comparable hemodynamic impairment was observed in the ophthalmic artery. Peripapillary choroidal thickness was bilaterally reduced in the inferior and temporal quadrants in both patient groups (p = 0.003 and p = 0.010). No significant difference was detected between affected and contralateral eyes in any parameter. RNFL, GC+, and macular thickness remained comparable across all groups. Conclusions: The bilateral symmetry of hemodynamic impairment and choroidal thinning suggests that SSNHL arises against a background of systemic microvascular disease. The combined use of OCT and color Doppler ultrasonography holds clinical potential as a non-invasive biomarker panel for defining the vascular phenotype of the condition. Full article

Recent advances in satellite technology have significantly progressed, yet acquiring high-quality images with meaningful labels for Earth observation missions remains a costly and time-intensive process. Furthermore, captured scenes frequently exhibit defects such as misaligned color channels, extensive cloud cover, or repetitive patterns in similar environments. Fortunately, the evolution of generative artificial intelligence offers a solution by enabling the creation of realistic synthetic scenes, simulating the characteristics of any targeted imager, and thereby mitigating the scarcity of authentic data. This paper demonstrates the feasibility of transferring knowledge from specialized AI-generated datasets to Earth observation missions. Leveraging a novel dataset of Spanish map tiles, Pix2Pix, CUT, and ControlNet models were implemented to synthesize satellite imagery. To analyze structural and topological generalizability, identical U-Net instances were trained on the resulting collections for building, road, and water segmentation tasks, and subsequently tested on independent authentic imagery. The results reveal a clear decoupling between visual realism and functional utility. Incorporating synthetic samples into hybridized training datasets successfully surpassed the limitations of using real data alone, increasing maximum Dice scores by 0.9% (to 54.1% for buildings), 2.3% (to 38.6% for roads), and 4.1% (to 46.5% for waterbodies). This systematic validation establishes structural-guided synthetic data augmentation as a robust, adaptable strategy for Earth observation applications across diverse sensors and geometric objectives. Full article

We present the EGIDE (Edge-on Galaxies in the DESI survey) project—a catalog of 149,215 edge-on galaxy candidates created using the data of the DESI Legacy Imaging Survey DR10 images. The catalog size is ten times greater than its predecessor and covers more than half of the sky. It is constructed in an automatic way, utilizing the full power of manual annotations from the GalaxyZoo volunteers, implemented in the Zoobot neural model, which was fine-tuned to search for edge-on galaxies specifically. To ensure the credibility of the dataset, subsequent manual supervision was performed. The EGIDE catalog provides homogeneous SExtractor photometry in the $griz$ bands, total stellar mass estimates, redshifts for 98% of the sample, star formation rates, and other information. All of this is publicly available at The Edge-on Galaxy Database site. The preliminary analysis focused on differences between edge-on galaxies in the so-called blue sequence and red cloud populations. These galaxies demonstrate distinct properties: the number of redder galaxies decreases with increasing $a/b$ ratio faster than that of the bluer galaxies; galaxy thickness varies with galaxy color: red sequence galaxies are thicker than blue cloud galaxies; the flattening ratio $q=b/a$ increases significantly with total stellar mass $M_{★}$ only among redder cloud galaxies. It is an intriguing result that the same trend of q increasing at the high-mass end is detected by both the statistical models of figures of revolution and direct observations of edge-on galaxies in EGIDE independently. The full extent of this relationship’s validity can only be determined after properly accounting for the contributions of the bulge and the PSF. Full article

For modern agricultural management, the accuracy of plant identification is crucial. However, the task becomes challenging because crops and weeds at early growth stages often exhibit similar color, leaf morphology, and texture in two-dimensional images captured under field conditions, despite their clear biological differences in terms of botanical species, root systems, and phenological characteristics. Furthermore, computing hardware in the field also has strict limits. Therefore, we developed the RT-BMTR network to handle these physical constraints. Within this architecture, image data is processed through a bilateral hybrid backbone named Bi-HMB. The DSFM captures small local details, and MambaVision understands the broader background information. Then, these features are fused by RepNCSPELAN4. We adopted this structure to reduce redundant calculations. Next, the model determines its bounding boxes using the Inner-ShapeIoU loss function. This geometric constraint improves the detection of small targets. When evaluated on the CropAndWeed dataset, our model achieved an average precision (AP) at IoU threshold 0.5 (AP⁵⁰) of 68.1%, AP⁷⁵ of 54.8%, and a mean AP averaged over IoU thresholds from 0.5 to 0.95 (AP^50–95) of 50.9%. Detection precision recorded 26.5% for small objects, 44.7% for moderate ones, and with 59.3% for large objects. Rates for the first two categories saw enhancements of 16.2% and 4.6%. Overall, our modified model outperforms the original RT-DETR baseline. We also shrank the overall parameter count by 30.1%, alongside a 4.2% decrease in computational demand. Full article

Previous research has primarily focused on the restorative effects of environments on the general population, often overlooking the specific restorative capacity of urban settings for the disabled population. There is a lack of comprehensive investigation into the interaction between accessibility elements and urban restorativeness. This study, conducted in Shenzhen, Guangdong Province, China, categorizes streetscape accessibility elements for the disabled population and develops a recognition system based on an enhanced DeeplabV3+ framework. Semantic segmentation of streetscape accessibility elements was performed using 201,860 sampling points and 807,440 street view images. This study employed a combination of TrueSkill scoring, sentiment semantic analysis, LDA topic modeling, and LAB color clustering to quantify and visualize urban restorativeness. The impact of accessibility elements on urban restorativeness was explored using the XGBoost-SHAP model. Results indicate significant effects of architectural space constraints and high-density motor vehicle distribution on the safety of the disabled population’s mobility. The low pixel ratio of accessibility facilities and signs indicates insufficient infrastructure, while high landscape recognition rates exhibit significant spatial coverage heterogeneity. Detection rates for the disabled population in street views are nearly zero, highlighting a severe lack of inclusivity in pedestrian environments. Urban restorativeness exhibited a pattern of being higher in the south and east, and lower in the north and west. Among the accessibility elements, public green spaces (PGS) contributed the most to urban restorativeness, accounting for 25% of the impact, and the study elucidates the mechanisms through which various elements affect urban restorativeness. This absence stems from spatial competition, missing co-design, threshold effect conflicts, and color interference mechanisms. This research breaks away from traditional linear analytical frameworks and reveals the complex non-linear relationship between accessibility elements and urban restorativeness through the XGBoost-SHAP model, providing a quantitative decision-making tool for planning accessible environments in high-density cities. Full article

Reliable individual identification is essential for precision livestock farming. To reduce the performance disparity among coat-color groups in open-set sheep face recognition, this study proposes a color-adaptive recognition method. A coat color classifier routes images to dedicated recognition models, each equipped with a tailored preprocessing pipeline. The method introduces a Retinex-Guided Gradient Perception Convolution (RGPConv), which embeds Retinex physical priors into the convolution operator and generates spatially adaptive modulation maps from the illumination component to dynamically control gradient enhancement intensity, enabling illumination-aware adaptive gradient feature extraction. A Multi-scale Channel Attention Fusion (MCAF) mechanism is further designed to integrate multi-level semantic information. Experiments on the LSFW dataset show that the proposed method achieves an open-set recognition accuracy of 96.08%, outperforming the baseline Li-SheepFaceNet by 1.50 percentage points. Compared with the corresponding routing-only subset baselines, accuracy improves by 2.41 and 2.00 percentage points for black-faced and white-faced sheep, respectively. On the cross-scene RealSheepFace and cross-species GoatFace datasets, accuracy improves by 4.26 and 2.00 percentage points, suggesting potential cross-domain transferability. Full article

Coral reefs support marine biodiversity, fisheries, tourism, and coastal protection, but they are increasingly threatened by environmental stress and bleaching. Satellite-based reef monitoring has mainly relied on thermal metrics, especially Degree Heating Weeks (DHW), to represent bleaching risk. However, thermal exposure alone may not fully describe reef stress in optically complex coastal waters, where light availability, water clarity, and water-quality conditions can modify coral response. This limitation is important in Indonesia, where reefs span diverse coastal environments and many bleaching observations occur under relatively low DHW. In this study, we develop the Coral Reef Environmental Stress Index (CRESI), implemented as CRESI-Mamba, to estimate coral reef stress in Indonesia as a continuous and interpretable satellite-based stress index. CRESI-Mamba uses 26-week sequences of thermal variables from NOAA Coral Reef Watch and ocean-color variables from NASA Visible Infrared Imaging Radiometer Suite (VIIRS). The model decomposes the inferred stress into thermal, optical, and water-quality pathways, and maps the resulting stress index to bleaching probability for event-based evaluation. CRESI-Mamba was trained and evaluated using 8424 reef observations from eight Indonesian regions. In Leave-One-Region-Out cross-validation (LORO-CV), the model achieved a mean area under the receiver operating characteristic curve (AUC) of $0.795\pm 0.087$. In grouped 5-fold cross-validation, it achieved an AUC of $0.802\pm 0.024$, exceeding the DHW-only baseline ($0.627\pm 0.021$) and performing comparably to stronger thermal-only models, while providing a pathway-decomposed stress index. The estimated stress index separated bleached and not-bleached observations, with paired stress differences of $0.299\pm 0.098$ in LORO-CV and $0.281\pm 0.032$ in grouped 5-fold CV. Pathway analysis showed that the dominant stress pathway differed among regions, with optical stress dominant in several low-DHW bleaching cases. These results show that reef stress in Indonesia is better represented as a multi-pathway environmental condition than as thermal exposure alone. CRESI-Mamba provides a framework for interpreting satellite environmental histories as reef stress, while retaining bleaching probability as an evaluation output. Full article

The objective of white balance is to accurately estimate and subsequently eliminate the color of global illumination present in an image. Learning-based methods have gained prominence over statistical approaches due to their typically superior accuracy. However, these methods rely on high-quality, large datasets. The quality of a dataset is intrinsically tied to the volume of knowledge it encapsulates, specifically the uniformity in the distribution of labels. In this study, we expand the dataset by leveraging the camera imaging pipeline. Subsequently, we segment the image into 16 partially overlapping blocks that collectively encompass the entire image. We then propose a rudimentary neural network designed to train these blocks with consistent labels, yielding 16 predictive outcomes that serve as image features. These features are used to capture complex illumination and reflection data within the image. Utilizing these features, we employ a straightforward, fully connected neural network to calculate the color mapping function, thereby correcting the image colors. Experimental results show that the methodology proposed in this paper significantly surpasses existing state-of-the-art color constancy methods. Full article

Polarimetric color cameras are a forefront technology that simultaneously captures polarimetric and color information by analyzing polarization states across different color channels, commonly red, green, and blue. In general, each of these color channels can carry different polarization information. Therefore, measuring the polarization Stokes vector at several discrete wavelengths simultaneously and with the highest possible resolution is of interest in multiple research areas. However, when a commercial color polarization sensor is used under simultaneous narrowband RGB illumination mode, its channels cannot be assumed to represent independent wavelength channels. Spectral overlap of the color filters introduces color crosstalk between wavelength-dependent analyzer intensities, which may bias the reconstructed Stokes parameters if it is not corrected before polarimetric inversion. Several methods have been proposed in the literature to address the color crosstalk problem but they typically assume that the polarization state is identical for all wavelengths. This assumption does not generally hold for real samples, which exhibit wavelength-dependent depolarization, retardance, and dichroism. To the best of our knowledge, this is the first work presenting a method that addresses the color crosstalk problem without assuming that the polarization state is identical across all wavelengths. In addition, Fourier domain demosaicking techniques are applied to interpolate the data and reconstruct the images. The present study demonstrates how the proposed method leads to an accurate recovery of chromatic and polarimetric information on both synthetic and real-world datasets. To test our approach, narrowband light beams at three wavelengths (470, 554, 630 nm), with different spatial polarization and degree of linear polarization distributions, have been simulated and validated with simulated and experimental data. The results demonstrate the feasibility of the method for accurate three polarization channels measurements. Full article

Reliable delineation of brain tumor boundaries in multimodal magnetic resonance imaging (MRI) remains challenging despite substantial advances in deep learning–based segmentation. Although modern encoder–decoder architectures achieve strong volumetric overlap, precise geometric alignment of tumor contours remains inconsistent, particularly for small lesions and heterogeneous clinical cases. In neuro-oncology, even minor boundary deviations may influence surgical planning, radiotherapy targeting, and longitudinal treatment assessment. These limitations suggest that segmentation performance is not determined solely by network depth or loss design, but also by how multimodal information is structured prior to learning. We introduce CMYD-SurfaceNet, a scale-aware cascaded framework that restructures multimodal MRI inputs at the representation level to enhance boundary-sensitive segmentation. Rather than treating modalities as independently concatenated channels, selected sequences are first organized into a task-guided pseudo-RGB projection. This intermediate representation is subsequently transformed into the CMYK color space to disentangle shared luminance structure from modality-specific contrast dominance. To further encode geometric priors, a gradient-derived boundary density channel is incorporated to explicitly emphasize spatial discontinuities corresponding to tumor margins. The resulting CMYD representation is integrated within a two-stage nnU-Net cascade, where global tumor localization is followed by high-resolution region-of-interest refinement with auxiliary contour supervision. This scale-aware design improves sensitivity to small tumor components while stabilizing contour delineation. Extensive evaluation on the BraTS benchmark demonstrates consistent improvements in boundary-sensitive metrics. Compared with baseline nnU-Net, the proposed framework reduces HD95 from 3.6 mm to 2.4 mm and increases Surface Dice at 1 mm tolerance from 0.82 to 0.89, while maintaining competitive Dice performance. These findings suggest that representation-level structural decoupling, when combined with scale-aware refinement, may provide clinically relevant boundary-aware multimodal MRI segmentation support without increasing architectural complexity. Full article