Video Understanding

Researchers boost long-form video understanding efficiency by 6.89x and reduce memory usage by 84% using a novel learnable video token merging algorithm.

Training-free video motion control achieved via novel Motion Feature (MOFT) extraction from existing video diffusion models, offering architecture-agnostic insights and high performance.

VFIMamba uses state-space models for efficient and dynamic video frame interpolation, achieving state-of-the-art results by introducing a novel Mixed-SSM Block and curriculum learning.

TrackIME enhances video point tracking by cleverly pruning the search space, resulting in improved accuracy and efficiency.

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…

ConVRT: A novel framework restores turbulence-distorted videos by decoupling spatial and temporal information in a neural representation, achieving temporally consistent mitigation.

TAPTRv2 enhances point tracking by introducing an attention-based position update, eliminating cost-volume reliance for improved accuracy and efficiency.

SyncVIS: A new framework for video instance segmentation achieves state-of-the-art results by synchronously modeling video and frame-level information, overcoming limitations of asynchronous approache…

StreamFlow accelerates video optical flow estimation by 44% via a streamlined in-batch multi-frame pipeline and innovative spatiotemporal modeling, achieving state-of-the-art results.

Researchers introduce Video Gaussian Representation (VGR) for versatile video processing, embedding videos into explicit 3D Gaussians for intuitive motion and appearance modeling.

SlotSSMs: a novel framework for modular sequence modeling, achieving significant performance gains by incorporating independent mechanisms and sparse interactions into State Space Models.

Researchers developed SF-V, a single-step image-to-video generation model, achieving a 23x speedup compared to existing models without sacrificing quality, paving the way for real-time video synthesis…

ReVideo enables precise local video editing by independently controlling content and motion, overcoming limitations of existing methods and paving the way for advanced video manipulation.

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.

MM-Det, a novel algorithm, uses multimodal learning and spatiotemporal attention to detect diffusion-generated videos, achieving state-of-the-art performance on the new DVF dataset.

NVRC: A novel end-to-end neural video codec achieves 23% coding gain over VVC VTM by optimizing representation compression.

NeuroClips: groundbreaking fMRI-to-video reconstruction, achieving high-fidelity smooth video up to 6s at 8FPS by decoding both high-level semantics and low-level perception flows.

NaRCan: High-quality video editing via diffusion priors and hybrid deformation fields.

Multi-view Masked Contrastive Representation Learning (M²CRL) significantly boosts endoscopic video analysis by using a novel multi-view masking strategy and contrastive learning, achieving state-of-t…