English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recently segment anything model (SAM) has shown powerful segmentationcapability and has drawn great attention in computer vision fields. Massivefollowing works have developed various applications based on the pretrained SAMand achieved impressive performance on downstream vision tasks. However, SAMconsists of heavy architectures and requires massive computational capacity,which hinders the further application of SAM on computation constrained edgedevices. To this end, in this paper we propose a framework to obtain a tinysegment anything model (TinySAM) while maintaining the strong zero-shotperformance. We first propose a full-stage knowledge distillation method withonline hard prompt sampling strategy to distill a lightweight student model. Wealso adapt the post-training quantization to the promptable segmentation taskand further reduce the computational cost. Moreover, a hierarchical segmentingeverything strategy is proposed to accelerate the everything inference by$2\times$ with almost no performance degradation. With all these proposedmethods, our TinySAM leads to orders of magnitude computational reduction andpushes the envelope for efficient segment anything task. Extensive experimentson various zero-shot transfer tasks demonstrate the significantly advantageousperformance of our TinySAM against counterpart methods. Pre-trained models andcodes will be available at https://github.com/xinghaochen/TinySAM andhttps://gitee.com/mindspore/models/tree/master/research/cv/TinySAM.

TinySAM: Pushing the Envelope for Efficient Segment Anything Model