Recent advances in large multimodal models (LMMs) have enabled impressive reasoning and perception abilities, yet most existing training pipelines still depend on human-curated data or externally verified reward models, limiting their autonomy and scalability. In this work, we strive to improve LMM reasoning capabilities in a purely unsupervised fashion (without any annotated data or reward distillation). To this end, we propose a self-evolving framework, named EvoLMM, that instantiates two cooperative agents from a single backbone model: a Proposer, which generates diverse, image-grounded questions, and a Solver, which solves them through internal consistency, where learning proceeds through a continuous self-rewarding process. This dynamic feedback encourages both the generation of informative queries and the refinement of structured reasoning without relying on ground-truth or human judgments. When using the popular Qwen2.5-VL as the base model, our EvoLMM yields consistent gains upto ∼3% on multimodal math-reasoning benchmarks, including ChartQA, MathVista, and MathVision, using only raw training images. We hope our simple yet effective approach will serve as a solid baseline easing future research in self-improving LMMs in a fully-unsupervised fashion.

Underwater image restoration is essential for marine applications ranging from ecological monitoring to archaeological surveys, but effectively addressing the complex and spatially varying nature of underwater degradations remains a challenge. Existing methods typically apply uniform restoration strategies across the entire image, struggling to handle multiple co-occurring degradations that vary spatially and with water conditions. We introduce TIDE, a two stage inverse degradation estimation framework that explicitly models degradation characteristics and applies targeted restoration through specialized prior decomposition. Our approach disentangles the restoration process into multiple specialized hypotheses that are adaptively fused based on local degradation patterns, followed by a progressive refinement stage that corrects residual artifacts. Specifically, TIDE decomposes underwater degradations into four key factors, namely color distortion, haze, detail loss, and noise, and designs restoration experts specialized for each. By generating specialized restoration hypotheses, TIDE balances competing degradation factors and produces natural results even in highly degraded regions. Extensive experiments across both standard benchmarks and challenging turbid water conditions show that TIDE achieves competitive performance on reference based fidelity metrics while outperforming state of the art methods on non reference perceptual quality metrics, with strong improvements in color correction and contrast enhancement.

Accurate classification of computed tomography (CT) images is essential for diagnosis and treatment planning, but existing methods often struggle with the subtle and spatially diverse nature of pathological features. Current approaches typically process images uniformly, limiting their ability to detect localized abnormalities that require focused analysis. We introduce UGPL, an uncertainty-guided progressive learning framework that performs a global-to-local analysis by first identifying regions of diagnostic ambiguity and then conducting detailed examination of these critical areas. Our approach employs evidential deep learning to quantify predictive uncertainty, guiding the extraction of informative patches through a non-maximum suppression mechanism that maintains spatial diversity. This progressive refinement strategy, combined with an adaptive fusion mechanism, enables UGPL to integrate both contextual information and fine-grained details. Experiments across three CT datasets demonstrate that UGPL consistently outperforms state-of-the-art methods, achieving improvements of 3.29%, 2.46%, and 8.08% in accuracy for kidney abnormality, lung cancer, and COVID-19 detection, respectively. Our analysis shows that the uncertainty-guided component provides substantial benefits, with performance dramatically increasing when the full progressive learning pipeline is implemented.

To be released...

Histopathology image segmentation is essential for delineating tissue structures in skin cancer diagnostics, but modeling spatial context and inter-tissue relationships remains a challenge, especially in regions with overlapping or morphologically similar tissues. Current convolutional neural network (CNN)-based approaches operate primarily on visual texture, often treating tissues as independent regions and failing to encode biological context. To this end, we introduce Neural Tissue Relation Modeling (NTRM), a novel segmentation framework that augments CNNs with a tissue-level graph neural network to model spatial and functional relationships across tissue types. NTRM constructs a graph over predicted regions, propagates contextual information via message passing, and refines segmentation through spatial projection. Unlike prior methods, NTRM explicitly encodes inter-tissue dependencies, enabling structurally coherent predictions in boundary-dense zones. On the benchmark Histopathology Non-Melanoma Skin Cancer Segmentation Dataset, NTRM outperforms state-of-the-art methods, achieving a robust Dice similarity coefficient that is 4.9\% to 31.25\% higher than the best-performing models among the evaluated approaches. Our experiments indicate that relational modeling offers a principled path toward more context-aware and interpretable histological segmentation, compared to local receptive-field architectures that lack tissue-level structural awareness.

Plant disease detection using computer vision has significantly advanced precision agriculture, but adapting to new disease variants with minimal labeled examples remains a challenge, particularly when regional factors influence symptom expression. Current approaches primarily rely on transfer learning or data augmentation techniques, which often fail to adequately capture the intricate visual characteristics specific to plant diseases from limited examples. We introduce SPROUT (Symptom-centric Prototypical Representation Optimization and Uncertainty-aware Tuning), a few-shot learning framework that leverages an attention-weighted prototype refinement mechanism to identify and emphasize the most informative disease features in support examples.

Radiographic imaging is crucial for diagnosing bone fractures, but the absence of reliable uncertainty measures in existing models complicates the interpretation of ambiguous cases, especially in complex or noisy datasets. Current deep learning methods for fracture diagnosis such as convolutional neural networks (CNNs) and Transformers have improved detection accuracy, but typically rely on deterministic outputs that fail to estimate predictive confidence or handle mislabeled samples. We introduce BONE-ULR, a Bayesian approach for bone fracture diagnosis that integrates adaptive label refinement and spatially-aware uncertainty estimation to enhance both reliability and interpretability. Unlike existing approaches which provide binary predictions without insight into their confidence or failure cases, BONE-ULR produces a predictive distribution via multiple stochastic forward passes, enabling effective quantification of epistemic uncertainty and identification of ambiguous regions. Additionally, we introduce a dynamic label refinement strategy that ranks training samples by entropy and excludes high-uncertainty bone X-rays from supervision, mitigating the impact of mislabels and ambiguous fracture patterns while improving representation learning. Extensive experimental analysis validates that our approach significantly improves classification accuracy and calibration, achieving an F1-score of 85.91% and an Expected Calibration Error (ECE) of 0.141, surpassing state-of-the-art (SOTA) methods in both performance and reliability.

Underwater images suffer from severe degradations, including color distortions, reduced visibility, and loss of structural details due to wavelength-dependent attenuation and scattering. Existing enhancement methods primarily focus on spatial-domain processing, neglecting the frequency domain's potential to capture global color distributions and long-range dependencies. To address these limitations, we propose FUSION, a dual-domain deep learning framework that jointly leverages spatial and frequency domain information. FUSION independently processes each RGB channel through multi-scale convolutional kernels and adaptive attention mechanisms in the spatial domain, while simultaneously extracting global structural information via FFT-based frequency attention. A Frequency Guided Fusion module integrates complementary features from both domains, followed by inter-channel fusion and adaptive channel recalibration to ensure balanced color distributions. Extensive experiments on benchmark datasets (UIEB, EUVP, SUIM-E) demonstrate that FUSION achieves state-of-the-art performance, consistently outperforming existing methods in reconstruction fidelity (highest PSNR of 23.717 dB and SSIM of 0.883 on UIEB), perceptual quality (lowest LPIPS of 0.112 on UIEB), and visual enhancement metrics (best UIQM of 3.414 on UIEB), while requiring significantly fewer parameters (0.28M) and lower computational complexity, demonstrating its suitability for real-time underwater imaging applications.

Neural Radiance Fields (NeRFs) have transformed novel view synthesis, but achieving accurate surface geometry remains challenging, especially with sparse views. Current approaches either require dense viewpoint sampling or produce inconsistent geometry due to their reliance on purely photometric supervision. We introduce PDE-NeRF, a physics-informed optimization framework that enforces geometric consistency through Partial Differential Equation (PDE)-constrained density gradients. Unlike existing methods that use auxiliary losses, our approach directly shapes the underlying density field by aligning spatial derivatives with ground-truth surface normals. We combine this with an efficient hash-based encoding scheme to enable high-fidelity reconstruction from sparse views. Our model achieves up to 8.51 dB higher PSNR and a 0.062 reduction in LPIPS over NeRF-based baselines, demonstrating superior perceptual quality. PDE-NeRF maintains consistent surface details, even in areas visible from only 30–40 views in a 360-degree capture, effectively addressing a key challenge in neural scene reconstruction. Experiments on diverse synthetic scenes demonstrate that our method achieves superior geometric accuracy and visual quality, outperforming existing approaches across both 360° inward-facing and LLFF datasets.

Vision Transformers (ViTs) have redefined image classification by leveraging self-attention to capture complex patterns and long-range dependencies between image patches. However, a key challenge for ViTs is efficiently incorporating multi-scale feature representations, which is inherent in convolutional neural networks (CNNs) through their hierarchical structure. Graph transformers have made strides in addressing this by leveraging graph-based modeling, but they often lose or insufficiently represent spatial hierarchies, especially since redundant or less relevant areas dilute the image's contextual representation. To bridge this gap, we propose SAG-ViT, a Scale-Aware Graph Attention ViT that integrates multi-scale feature capabilities of CNNs, representational power of ViTs, and graph-attended patching to enable richer contextual representation. Using EfficientNetV2 as a backbone, the model extracts multi-scale feature maps, dividing them into patches to preserve richer semantic information compared to directly patching the input images. The patches are structured into a graph using spatial and feature similarities, where a Graph Attention Network (GAT) refines the node embeddings. This refined graph representation is then processed by a Transformer encoder, capturing long-range dependencies and complex interactions. We evaluate SAG-ViT on benchmark datasets across various domains, validating its effectiveness in advancing image classification tasks. Our code and weights are available at https://github.com/shravan-18/SAG-ViT.

Neurodegenerative (ND) diseases are autoimmune diseases that affect the central nervous system, including the brain and spinal cord. In recent years, deep learning has demonstrated its potential in medical imaging for diagnostic purposes. However, for these techniques to be fully accepted in clinical settings, they must achieve high performance and gain the confidence of medical professionals regarding their interpretability. Therefore, an interpretable model should make decisions based on clinically relevant information like a domain expert. To achieve this, we present an interpretable classifier dedicated to the most common autoimmune ND diseases. The lesions associated with ND diseases exhibit irregular distributions and spatial dependencies in different regions of the brain, challenging traditional models to effectively capture both local and global relationships. To address this issue, we present a Residual Graph Neural Network enhanced Vision Transformer (RG-ViT) that represents MRI data as a graph of interconnected patches. By integrating residual connections into the GNN framework, we preserve critical features while promoting effective message passing. This approach overcomes the problem of spatial disconnection prevalent in standard patch-based methods and provides a cohesive and context-aware analysis of MRI data. Experimental results in detecting multiple sclerosis (MS), Parkinson's (PD), and Alzheimer's disease (AD) demonstrated our approach's consistent accuracy scores of 98.7%, 99.6%, and 99.1%, respectively. On the combined dataset for the global classification of ND diseases, it achieved an F1 score of 99.2%, justifying its generalizability.

Tumors in the brain are caused by abnormal growths in brain tissue resulting from different types of brain cells. If undiagnosed, they lead to aggressive neurological deficits, including cognitive impairment, motor dysfunction, and sensory loss. With the growth of the tumor, intracranial pressure will definitely increase, and this may bring about such dramatic complications as herniation of the brain, which may be fatal. Hence, early diagnosis and treatment are required to control the complications arising due to such tumors to retard their growth. Several works related to deep learning (DL) and artificial intelligence (AI) are being conducted to help doctors diagnose at an early stage by using the scans taken from Magnetic Resonance Imaging (MRI). Our research proposes targeted neural architectures within multi-objective frameworks that can localize, segment, and classify the grade of these gliomas from multimodal MRI images to solve this critical issue. Our localization framework utilizes a targeted architecture that enhances the LinkNet framework with an encoder inspired by VGG19 for better multimodal feature extraction from the tumor along with spatial and graph attention mechanisms that sharpen feature focus and inter-feature relationships. For the segmentation objective, we deployed a specialized framework using the SeResNet101 CNN model as the encoder backbone integrated into the LinkNet architecture, achieving an IoU Score of 96%. The classification objective is addressed through a distinct framework implemented by combining the SeResNet152 feature extractor with Adaptive Boosting classifier, reaching an accuracy of 98.53%. Our multi-objective approach with targeted neural architectures demonstrated promising results for complete glioma characterization, with the potential to advance medical AI by enabling early diagnosis and providing more accurate treatment options for patients.

Understanding emotions is a fundamental aspect of human communication. Integrating audio and video signals offers a more comprehensive understanding of emotional states compared to traditional methods that rely on a single data source, such as speech or facial expressions. Despite its potential, multimodal emotion recognition faces significant challenges, particularly in synchronization, feature extraction, and fusion of diverse data sources. To address these issues, this paper introduces a novel transformer-based model named Audio-Video Transformer Fusion with Cross Attention (AVT-CA). The AVT-CA model employs a transformer fusion approach to effectively capture and synchronize interlinked features from both audio and video inputs, thereby resolving synchronization problems. Additionally, the Cross Attention mechanism within AVT-CA selectively extracts and emphasizes critical features while discarding irrelevant ones from both modalities, addressing feature extraction and fusion challenges. Extensive experimental analysis conducted on the CMU-MOSEI, RAVDESS and CREMA-D datasets demonstrates the efficacy of the proposed model. The results underscore the importance of AVT-CA in developing precise and reliable multimodal emotion recognition systems for practical applications.

Autonomous vehicular technology, also known as self-driving or driverless technology, refers to the innovation that enables vehicles to operate without human intervention. Traffic sign classification (TSC) is a critical component in autonomous vehicular technology, as it allows vehicles to recognize and interpret traffic signs, which is essential for safe and rule-compliant navigation. This work proposes a novel attention-fused deep convolutional neural network (AFDCNN) for TSC. The proposed AFDCNN incorporates the capabilities of ResNet50 and EfficientNetV2 by combining their outputs through a self-attention mechanism which enhances its ability to classify traffic signs. Analysis of the GTSRB, LISA, and MASTIF datasets revealed that the proposed model exhibited superior performance compared to state-of-the-art models, as evidenced by higher scores in recall, precision, F1-score, and accuracy metrics.

Tomatoes are one of the most widely consumed and economically significant food crops globally, with their yield quality and quantity significantly impacted by various diseases. Early disease detection is crucial to reduce their effects and improve crop yields, supporting farmers. While previous research has applied machine learning techniques to segment and classify tomato leaf images, existing classifiers often struggle with accurately detecting new disease types. This study proposes a novel approach to tomato leaf disease classification by harnessing the power of deep learning and swarm intelligence-based optimization techniques. An ensemble model is developed, integrating an exponential moving average function with temporal constraints and an enhanced weighted gradient optimizer into fine-tuned Visual Geometry Group-16 (VGG-16) and Neural Architecture Search Network (NASNet) mobile training methods. The model is trained and validated on a dataset of 10,000 tomato leaf images categorized into nine disease classes, with an additional 1,000 images reserved for testing. The results show superior performance across key metrics: accuracy (98.7%), loss (4%), precision (97.9%), recall (98.6%), receiver operating characteristic curve (99.97%), and F1-score (98.7%), outperforming existing methods and enhancing disease detection accuracy.