TinySAM: Pushing the Envelope for Efficient Segment Anything Model
Proceedings of the AAAI Conference on Artificial Intelligence · 11.04.2025 · 39 (19) · 20470-20478
DOI: 10.1609/aaai.v39i19.34255 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 11.04.2025. Verfasst von Han Shu und 7 weiteren Autor:innen. Die Arbeit wurde bisher 44-mal zitiert (Crossref).
Abstract (Originalsprache)
Recently segment anything model (SAM) has shown powerful segmentation capability and has drawn great attention in computer vision fields. Massive following works have developed various applications based on the pre-trained SAM and achieved impressive performance on downstream vision tasks. However, SAM consists of heavy architectures and requires massive computational capacity, which hinders the further application of SAM on computation constrained edge devices. To this end, in this paper we propose a framework to obtain a tiny segment anything model (TinySAM) while maintaining the strong zero-shot performance. We first propose a full-stage knowledge distillation method with hard prompt sampling and hard mask weighting strategy to distill a lightweight student model. We also adapt the post-training quantization to the prompt-based segmentation task and further reduce the computational cost. Moreover, a hierarchical segmenting everything strategy is proposed to accelerate the everything inference by 2× with almost no performance degradation. With all these proposed methods, our TinySAM leads to orders of magnitude computational reduction and pushes the envelope for efficient segment anything task. Extensive experiments on various zero-shot transfer tasks demonstrate the significantly advantageous performance of our TinySAM against counterpart methods.
Zitieren
APA
Shu, H., Li, W., Tang, Y., Zhang, Y., Chen, Y., Li, H., Wang, Y., & Chen, X. (2025). TinySAM: Pushing the Envelope for Efficient Segment Anything Model. Proceedings of the AAAI Conference on Artificial Intelligence, 39(19), 20470-20478. https://doi.org/10.1609/aaai.v39i19.34255
Vancouver
Shu H, Li W, Tang Y, Zhang Y, Chen Y, Li H, et al. TinySAM: Pushing the Envelope for Efficient Segment Anything Model. Proceedings of the AAAI Conference on Artificial Intelligence. 2025;39(19):20470-20478. doi:10.1609/aaai.v39i19.34255
BibTeX
@article{Shu2025,
author = {Shu, Han and Li, Wenshuo and Tang, Yehui and Zhang, Yiman and Chen, Yihao and Li, Houqiang and Wang, Yunhe and Chen, Xinghao},
title = {TinySAM: Pushing the Envelope for Efficient Segment Anything Model},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2025},
volume = {39},
number = {19},
pages = {20470--20478},
doi = {10.1609/aaai.v39i19.34255}
}
EUVIMED