ant0071/samurai

SAMURAI is a zero-shot visual tracking system that adapts Meta's Segment Anything Model (SAM 2) for object tracking by incorporating motion-aware memory, enabling it to track objects in videos without any training on tracking-specific data. Developed by researchers at the University of Washington's Information Processing Lab, it achieves state-of-the-art performance across major visual object tracking benchmarks including LaSOT, GOT-10k, and TrackingNet. The project is intended for computer vision researchers and developers working on video analysis and object tracking applications.