# IDEA-Research/Grounded-SAM-2

Grounded SAM 2: Ground and Track Anything in Videos with Grounding DINO, Florence-2 and SAM 2

Repository: https://github.com/IDEA-Research/Grounded-SAM-2
Canonical: https://ross.abutalabs.com/products/grounded-sam-2
Homepage: https://arxiv.org/abs/2401.14159
Language: Jupyter Notebook
License: Apache-2.0
License Family: permissive
Last push: 2025-11-11T12:16:57+00:00

## Health v2 (maintenance only)
Score: 37/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 51, release rhythm 8, longevity 54
- inputs: {"age_days": 762, "days_push": 295, "days_rel": 630, "gap_med": null, "n_releases_24m": 1}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3708, forks 430 (observed 2026-08-28T04:08:15.139895+00:00)

## What it is
Grounded SAM 2 is a foundation-model pipeline that combines open-set detectors (Grounding DINO, Grounding DINO 1.5/1.6, Florence-2, DINO-X) with SAM 2 to detect, segment, and track arbitrary objects in images and videos from text prompts. It provides simple demo implementations and visualization built on the supervision library.

## Use cases
- segment any object in an image using a text prompt
- track objects across video frames with text-based grounding
- auto-annotate datasets with masks for training segmentation models
- detect and segment dense small objects in high-resolution 4K images with SAHI
- visualize detection, segmentation, and tracking results on videos
- run zero-shot open-vocabulary segmentation benchmarks

## When to choose
- you need text-prompted open-vocabulary detection and segmentation without training a custom model
- you want to segment and track objects in videos with a proven research pipeline
- you need automated mask annotation for dataset labeling
- you want simple reference implementations combining Grounding DINO/Florence-2/DINO-X with SAM 2

## When to avoid
- you need a production-ready optimized inference service rather than demo notebooks
- you lack GPU resources, since the models are heavy to run locally
- you only need classic closed-set detection with fixed classes, where a lightweight YOLO-style detector suffices
- you need real-time performance on edge devices

## Facets
- artifact type: library
- maturity: active
- function: computer-vision, image-processing, machine-learning, deep-learning
- domain: computer-vision, image-processing, artificial-intelligence, deep-learning
- platform: python, cross-platform
- tags: object-detection, image-segmentation, video-object-tracking, open-vocabulary-detection, sam2, grounding-dino, florence-2, dino-x, zero-shot, jupyter-notebooks, video, gpu

## Member repositories
- IDEA-Research/Grounded-SAM-2 (main) score 37

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:15.139895+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 2026-08-29T18:30:56.717472+00:00, confidence not recorded.
  - readme: https://github.com/IDEA-Research/Grounded-SAM-2 (fetched 2026-08-28T04:08:15.139895+00:00, sha b8084bb7784d)
  - homepage: https://arxiv.org/abs/2401.14159 (fetched 2026-08-29T09:24:29.894572+00:00, sha 9a6bd885f6d5)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T09:24:29.906607+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T09:24:29.903472+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T09:24:29.908448+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T09:24:29.905094+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.