# ZrrSkywalker/Personalize-SAM

Personalize Segment Anything Model (SAM) with 1 shot in 10 seconds

Repository: https://github.com/ZrrSkywalker/Personalize-SAM
Canonical: https://ross.abutalabs.com/products/personalize-sam
Language: Python
License: MIT
License Family: permissive
Last push: 2024-07-22T15:13:43+00:00

## Health v2 (maintenance only)
Score: 29/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 86
- inputs: {"age_days": 1217, "days_push": 772, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1671, forks 113 (observed 2026-08-28T04:05:20.109198+00:00)

## What it is
PerSAM is the official implementation of 'Personalize Segment Anything Model with One Shot', which customizes the Segment Anything Model (SAM) to segment specific visual concepts (e.g., a pet dog) in other images or videos given a single reference image with a mask. It offers a training-free variant and a fine-tuned variant (PerSAM-F) that trains only 2 parameters in 10 seconds.

## Use cases
- segment my pet dog across a whole photo album from one labeled example
- personalize SAM for one-shot object segmentation without training
- segment a specific object in videos given a single reference mask
- remove background disturbance for DreamBooth fine-tuning of Stable Diffusion
- fine-tune SAM on my own dataset with a single shot in seconds
- run a web demo of personalized segmentation

## When to choose
- you need to segment a specific object instance across many images or videos with only one annotated example
- you want a training-free or near-instant (10-second) SAM personalization
- you want to clean up few-shot training images for DreamBooth/Stable Diffusion personalization
- you need efficient segmentation on resource-constrained setups via MobileSAM support

## When to avoid
- you need general-purpose segmentation of arbitrary objects without a reference mask
- you need multi-shot or large-scale supervised segmentation training
- you need a production-ready application with a polished UI rather than a research codebase
- you work outside Python/GPU environments

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, computer-vision, image-processing
- domain: computer-vision, image-processing, deep-learning, artificial-intelligence
- platform: python
- tags: segment-anything, sam, one-shot-segmentation, personalized-segmentation, persam, image-segmentation, training-free, stable-diffusion-assist, gpu

## Member repositories
- ZrrSkywalker/Personalize-SAM (main) score 29

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:20.109198+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 2026-08-30T03:42:40.479607+00:00, confidence not recorded.
  - readme: https://github.com/ZrrSkywalker/Personalize-SAM (fetched 2026-08-28T04:05:20.109198+00:00, sha 8f7fbde121ee)
- Data as of 2026-08-30T08:39:29.467469+00:00.