Computer Vision

Training-free guidance revolutionizes diffusion models by enabling zero-shot conditional generation, but suffers from misaligned gradients and slow convergence. This paper provides theoretical analysi…

UMB: A novel model enhances open-world object detection by understanding model behavior, surpassing state-of-the-art with a 5.3 mAP gain for unknown classes.

UltraPixel generates high-quality images at various resolutions (1K-6K) efficiently using cascade diffusion models, achieving state-of-the-art performance.

UDPM: Upsampling Diffusion Probabilistic Models achieves high-quality image generation with fewer computations by incorporating downsampling and upsampling within the diffusion process.

UDON: a novel multi-teacher online distillation method creates highly efficient universal image embeddings by dynamically transferring domain-specific knowledge and adapting to imbalanced data.

U-DiT: Revolutionizing diffusion transformers with a U-Net design and token downsampling for superior image generation and drastically reduced computation cost.

Typicalness-Aware Learning (TAL) improves failure detection by dynamically adjusting prediction confidence based on sample typicality, mitigating overconfidence and achieving significant performance g…

Efficient Multi-Task Learning (EMTAL) transforms pre-trained Vision Transformers into efficient multi-task learners by using a MoEfied LoRA structure, a Quality Retaining optimization, and a router fa…

Transformer Doctor diagnoses and treats vision transformer errors by identifying and correcting information integration issues, improving model performance and interpretability.

UMI-GRAT: A universal meta-initialized and gradient robust adversarial attack effectively exploits vulnerabilities in the Segment Anything Model (SAM) and its fine-tuned downstream models, even withou…

AdaptiveDiffusion accelerates diffusion model inference by adaptively skipping noise prediction steps, achieving 2-5x speedup without quality loss.

Train open-vocabulary 3D object detectors using only RGB images and large language models, achieving state-of-the-art performance without expensive LiDAR data.

TrAct boosts vision model training by directly optimizing first-layer activations, leading to significant speedups (1.25x-4x) and improved accuracy.

Boosting diffusion-based human image animation, Test-time Procrustes Calibration (TPC) ensures high-quality outputs by aligning reference and target images, overcoming common compositional misalignmen…

Unsupervised model selection for domain adaptive object detection is achieved via a new Detection Adaptation Score (DAS), effectively selecting optimal models without target labels by leveraging the f…

PixelCLIP: Open-vocabulary semantic segmentation without pixel-level labels! Leveraging unlabeled image masks from Vision Foundation Models and an online clustering algorithm, PixelCLIP achieves imp…

Researchers propose Multi-Domain learning for Video Anomaly Detection (MDVAD) to create generalizable models handling conflicting abnormality criteria across diverse datasets, improving accuracy and a…

GroupEXP-DA boosts domain adaptive 3D object detection by using a grouping-exploration strategy to reduce bias in pseudo-label collection and account for multiple factors affecting object perception i…

Partial2Global: A novel VICL framework achieving globally optimal prompt selection, significantly improving visual in-context learning across various tasks.

FleVRS: One unified model masters standard, promptable, and open-vocabulary visual relationship segmentation, outperforming existing methods.