# chongzhou96/EdgeSAM

Official PyTorch implementation of "EdgeSAM: Prompt-In-the-Loop Distillation for On-Device Deployment of SAM"

Repository: https://github.com/chongzhou96/EdgeSAM
Canonical: https://ross.abutalabs.com/products/edgesam
Homepage: https://mmlab-ntu.com/project/edgesam/
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Topics: on-device-ai, segment-anything, coreml
Last push: 2025-05-24T07:54:29+00:00

## Health v2 (maintenance only)
Score: 37/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 23, release rhythm 35, longevity 71
- inputs: {"age_days": 1001, "days_push": 466, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1173, forks 61 (observed 2026-08-28T04:03:51.926669+00:00)

## What it is
EdgeSAM is the official PyTorch implementation of a distilled, accelerated variant of the Segment Anything Model (SAM) designed for on-device deployment. It distills SAM's ViT encoder into a CNN-based architecture with prompts in the loop, achieving 40x speedup over SAM and over 30 FPS on an iPhone 14.

## Use cases
- run segment anything model on mobile devices
- fast interactive image segmentation with box and point prompts
- distill SAM into a lightweight CNN model
- deploy segmentation model on iPhone with CoreML
- speed up SAM inference on edge devices
- annotate images with a fast SAM variant
- export SAM-like model to ONNX for web demos

## When to choose
- you need SAM-quality promptable segmentation on resource-constrained or mobile hardware
- you want faster inference than MobileSAM with better accuracy
- you are deploying interactive segmentation on iOS or edge devices

## When to avoid
- you need the full accuracy of the original ViT-based SAM on server hardware
- you need segmentation without box or point prompts
- you need a non-research license for commercial use

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, computer-vision, image-processing
- domain: computer-vision, image-processing, machine-learning, artificial-intelligence
- platform: python, cross-platform
- tags: segment-anything, knowledge-distillation, on-device-ai, coreml, onnx, sam, interactive-segmentation, edge-devices, ios

## Member repositories
- chongzhou96/EdgeSAM (main) score 37

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:51.926669+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:28:21.829174+00:00, confidence not recorded.
  - readme: https://github.com/chongzhou96/EdgeSAM (fetched 2026-08-28T04:03:51.926669+00:00, sha 3e79ec7b96c2)
  - homepage: https://mmlab-ntu.com/project/edgesam/ (fetched 2026-08-29T12:33:50.558128+00:00, sha abd0d00d5bac)
- Data as of 2026-08-30T08:39:29.467469+00:00.