Computer Vision

PeRFlow accelerates diffusion models by straightening their sampling trajectories using a piecewise reflow operation, enabling fast and high-quality image generation with minimal computational cost.

BiXT, a novel bi-directional cross-attention Transformer, scales linearly with input size, achieving competitive performance across various tasks by efficiently processing longer sequences.

New Pedestrian-Vehicle Collision Pose dataset (PVCP) and Pose Estimation Network (PPSENet) improve pedestrian pre-collision pose estimation from dashcam video.

PCoTTA: A novel framework enables multi-task point cloud models to seamlessly adapt to continuously changing target domains during testing, overcoming catastrophic forgetting and error accumulation.

AdaptIR: A novel parameter-efficient method for generalized image restoration using a heterogeneous Mixture-of-Experts (MoE) architecture that achieves superior performance and generalization.

PaGoDA: Train high-resolution image generators efficiently by progressively growing a one-step generator from a low-resolution diffusion model. This innovative pipeline drastically cuts training cost…

OPUS: a novel, real-time occupancy prediction framework using a sparse set prediction paradigm, outperforms state-of-the-art methods on Occ3D-nuScenes.

OSDCap: Online optimal-state dynamics estimation selectively incorporates physics models with kinematic observations to achieve highly accurate, physically-plausible human motion capture from videos.

Optimal Transport-based Labor-free Text Prompt Modeling (OLTM) leverages VQA and optimal transport for highly accurate sketch-based person re-identification without manual labeling.

Researchers created an energy-efficient optical system for generating images using light propagation, drastically reducing the latency and energy consumption of diffusion models.

OpenGaussian achieves 3D point-level open vocabulary understanding using 3D Gaussian Splatting by training 3D instance features with high 3D consistency, employing a two-level codebook for feature dis…

OpenDlign uses novel depth-aligned images from a diffusion model to boost open-world 3D understanding, achieving significant performance gains on diverse benchmarks.

Language Hierarchical Self-training (LHST) enhances weakly-supervised object detection by integrating language hierarchy, mitigating label mismatch, and improving generalization across diverse dataset…

OPEL: a novel optimal transport framework for procedure learning, significantly outperforms SOTA methods by aligning similar video frames and relaxing strict temporal assumptions.

OnlineTAS, a novel framework, achieves state-of-the-art performance in online temporal action segmentation by using an adaptive memory and a post-processing method to mitigate over-segmentation.

OneActor: One-shot tuning for consistent subject image generation, bypassing laborious backbone tuning via semantic guidance, achieving 4x faster speed.

One-to-Normal: Anomaly personalization boosts few-shot anomaly detection accuracy by transforming query images to match normal data, enabling precise, robust comparisons and flexible integration with …

PSTUDA, a novel progressive style transfer framework, efficiently segments kidney tumors across multiple MRI sequences using unsupervised domain adaptation, achieving higher accuracy and efficiency th…

OSEDiff: One-step diffusion network for real-world image super-resolution, achieving comparable or better results than multi-step methods with significantly reduced computational cost and improved ima…

Score Implicit Matching (SIM) revolutionizes diffusion model distillation by creating high-quality, single-step generators from complex, multi-step models, achieving comparable performance and enablin…