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

This paper presents EdgeSAM, an accelerated variant of the Segment AnythingModel (SAM), optimized for efficient execution on edge devices with minimalcompromise in performance. Our approach involves distilling the originalViT-based SAM image encoder into a purely CNN-based architecture, better suitedfor edge devices. We carefully benchmark various distillation strategies anddemonstrate that task-agnostic encoder distillation fails to capture the fullknowledge embodied in SAM. To overcome this bottleneck, we include both theprompt encoder and mask decoder in the distillation process, with box and pointprompts in the loop, so that the distilled model can accurately capture theintricate dynamics between user input and mask generation. To mitigate datasetbias issues stemming from point prompt distillation, we incorporate alightweight module within the encoder. EdgeSAM achieves a 40-fold speedincrease compared to the original SAM, and it also outperforms MobileSAM, being14 times as fast when deployed on edge devices while enhancing the mIoUs onCOCO and LVIS by 2.3 and 3.2 respectively. It is also the first SAM variantthat can run at over 30 FPS on an iPhone 14. Code and models are available athttps://github.com/chongzhou96/EdgeSAM.

EdgeSAM: Prompt-In-the-Loop Distillation for On-Device Deployment of SAM