SAMURAI
lastmod 2024-11-25created 2024-11-252 edits
Official repository of "SAMURAI: Adapting Segment Anything Model for Zero-Shot Visual Tracking with Motion-Aware Memory"
Introduction
The motion tracking model "SAMURAI" has been released as open source.
Meta's existing SAM2 was already quite good, and now an even better model built on it has been released as open source.
SAMURAI automatically tracks the object you want in a video in real time, and it holds on to it better than SAM2!
- Improved motion awareness: accurately tracks objects even with many people or fast movement
- High benchmark scores: achieved top zero-shot performance on several benchmarks such as LaSOT and GOT-10k.
- Open source: released under the Apache 2.0 license, so free for commercial use
Looks like it could be useful for K-pop fancam editing, sports analysis, and more!!
Documentation
- [2411.11922] SAMURAI - Adapting Segment Anything Model for Zero-Shot Visual Tracking with Motion-Aware Memory
- https://arxiv.org/abs/2411.11922
See also
- Deep learning
- PyTorch
- detectron
- detectron2
- Mask R-CNN
- MediaPipe
- You Only Look At CoefficienTs (YOLACT)
- Sahi
- GAMMA
- mmdetection
- Animated Drawings
- Segment Anything Model 2 (SAM 2)