anuragxel/salt

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1048 stars · 126 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

SALT is a Python-based image labeling tool built on Meta AI's Segment Anything Model, providing a barebones GUI for annotating images with predicted segmentation masks. It saves annotations in COCO format and uses precomputed embeddings and ONNX models so the labeling machine does not need a GPU.

Use cases

• label images for instance segmentation datasets
• generate COCO format annotations for object detection training
• speed up image segmentation annotation with SAM
• annotate datasets without a GPU on the labeling machine
• create segmentation masks by clicking on objects

When to choose

• you need COCO-format segmentation annotations for training models
• you want SAM-assisted click-based labeling to speed up annotation
• your labeling machine lacks a GPU since inference runs elsewhere via embeddings/ONNX

When to avoid

• you need a polished, stable annotation UI - the tool is explicitly rough and under active development
• you need polygon/box-only annotation without segmentation masks
• you want a web-based or multi-user annotation platform

Facets

application · maturity active

image-processing computer-vision machine-learning gui computer-vision image-processing machine-learning developer-tools python cross-platform segment-anything image-annotation coco-format data-labeling onnx computer-vision-datasets desktop gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="anuragxel/salt")