The University of Washington has unveiled SAMURAI, a next-generation visual tracking model that enhances object tracking in complex scenes. Building on the Segment Anything Model2 (SAM2), SAMURAI addresses challenges like occlusion and fast-moving objects, achieving strong zero-shot performance across benchmark datasets.

The University of Washington has unveiled SAMURAI, a next-generation visual tracking model designed to improve tracking performance in complex scenes. Building on the Segment Anything Model2 (SAM2), SAMURAI addresses challenges such as fast-moving and self-occluding objects, showing remarkable zero-shot performance and enhanced accuracy across various benchmark datasets.