5563 papers
This paper introduces PhotoBench, the first benchmark constructed from authentic personal albums to shift photo retrieval from simple visual matching to complex, intent-driven reasoning by exposing critical limitations in current unified embedding and agentic systems regarding non-visual constraints and multi-source fusion.
This paper introduces a rate-distortion-theoretic framework that characterizes the generalization trade-offs of human and machine vision systems using geometric signatures of slope and curvature, revealing that while both follow a common lossy-compression principle, humans exhibit smoother and more flexible trade-offs compared to the steeper, more brittle regimes of modern deep networks.
This paper proposes a Downstream Task-Inspired Underwater Image Enhancement (DTI-UIE) framework that integrates a human visual perception model, a task-driven perceptual loss, and an automatically constructed dataset to generate enhanced images specifically optimized for improving downstream vision tasks like object detection and semantic segmentation.
This paper introduces Neural Operator-Grounded Continuous Tensor Function Representation (NO-CTR), a novel framework that replaces discrete, linear mode- products with continuous, nonlinear neural operators to more faithfully represent complex real-world data across various grid structures and point clouds, while theoretically guaranteeing universal approximation and demonstrating superior performance in multi-dimensional data completion tasks.
FireRed-OCR is a novel framework that transforms general-purpose VLMs into high-performance, pixel-precise document parsing experts by leveraging a "Geometry + Semantics" data factory and a three-stage progressive training strategy to overcome structural hallucinations and achieve state-of-the-art results on complex document benchmarks.
This paper introduces Tiny-DroNeRF, a lightweight Neural Radiance Field model optimized for sub-100mW microcontrollers on nano-drones and enhanced by a federated learning scheme, which achieves dense 3D scene reconstruction with a 96% memory reduction and improved accuracy despite extreme resource constraints.
This paper proposes a novel event-only drone trajectory forecasting method that extracts propeller rotational speed directly from raw event data and integrates it into an RPM-aware Kalman filter, achieving superior short-to-medium horizon prediction accuracy compared to learning-based approaches without relying on RGB imagery or training data.
This paper introduces a training-free, noise-robust 3D Field of Junctions (3D FoJ) representation that optimizes volumetric wedge junctions to serve as a structural prior, successfully outperforming both classical and neural methods in low-SNR 3D imaging tasks such as CT, cryo-ET, and point cloud denoising without risking hallucination.
This paper proposes a novel data augmentation method called Conditional CycleGAN Mixup Augmentation (C2GMA) that leverages visible-band imagery to synthesize mixed-class non-visible domain examples via CycleGANs, significantly improving classification accuracy in data-scarce Synthetic Aperture Radar (SAR) applications.
The paper introduces Dite-HRNet, a dynamic lightweight high-resolution network that incorporates novel dynamic split convolution and adaptive context modeling to efficiently capture multi-scale features and long-range spatial dependencies, achieving superior performance on human pose estimation benchmarks compared to state-of-the-art lightweight networks.
The paper proposes CO^3, a novel unsupervised framework for outdoor 3D point cloud representation learning that leverages cooperative vehicle- and infrastructure-side LiDAR views along with contextual shape prediction to overcome reconstruction challenges and achieve state-of-the-art performance on downstream detection tasks.
This paper proposes a straightforward and effective Convolutional Neural Network (CNN) scheme for detecting faults in PV solar panels, achieving high accuracy rates of 91.1% for binary classification and 88.6% for multi-classification while outperforming previous models using the same datasets.
This paper addresses reward overoptimization in diffusion models by identifying mismatches with temporal inductive bias and the regularization role of dormant neurons, leading to the proposal of TDPO-R, a policy gradient algorithm that leverages temporal inductive bias and resets active neurons to mitigate primacy bias.
This paper introduces Uni-ISP, a novel pipeline that unifies the learning of image signal processors for diverse mobile cameras using device-aware embeddings and a specialized training scheme, supported by a new real-world 4K dataset called FiveCam, to achieve superior accuracy, adaptability, and versatility in both forward and inverse ISP tasks.
This paper proposes R2GenCSR, a novel radiology report generation framework that leverages the linear-complexity Mamba architecture for efficient visual feature extraction and enhances LLM performance by mining both contextual and residual information from training samples to generate high-quality medical reports.
The paper proposes Shuffle Mamba, a novel multi-modal image fusion framework that employs a Bayesian-inspired Random Shuffle scanning strategy and its inverse to eliminate biases from fixed scanning patterns, thereby achieving robust cross-modality interaction and superior fusion quality through Monte-Carlo averaging.
This paper proposes the first label unlearning method for Vertical Federated Learning that utilizes a representation-level manifold mixup mechanism to generate synthetic embeddings for gradient-based forgetting and recovery, effectively removing sensitive label information while preserving model utility and computational efficiency across diverse datasets.
Stereo-Talker is a novel one-shot audio-driven system that synthesizes high-fidelity, 3D talking videos with precise lip synchronization and expressive gestures by leveraging LLM priors for motion generation and a prior-guided Mixture-of-Experts mechanism for stable, view-controllable video synthesis.
This paper proposes Stepwise Diffusion Policy Optimization (SDPO), a novel reinforcement learning framework that aligns few-step diffusion models with specific objectives by introducing dual-state trajectory sampling and latent similarity-based dense reward prediction to enable low-variance, granular policy updates and superior reward-aligned image synthesis.
TREND introduces the first unsupervised 3D representation learning framework for LiDAR perception that leverages temporal forecasting via a Recurrent Embedding scheme and Temporal Neural Field to predict future observations, achieving significant performance gains over existing methods on downstream 3D object detection tasks.