# graspnet/graspnet-baseline

Baseline model for "GraspNet-1Billion: A Large-Scale Benchmark for General Object Grasping" (CVPR 2020)

Repository: https://github.com/graspnet/graspnet-baseline
Canonical: https://ross.abutalabs.com/products/graspnet-baseline
Homepage: https://graspnet.net/
Language: Python
License: NOASSERTION
License Family: other
Topics: point-cloud, grasping, 3d, deep-learning, robotics, graspnet-1billion
Last push: 2025-02-17T17:54:09+00:00

## Health v2 (maintenance only)
Score: 35/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 7, release rhythm 35, longevity 100
- inputs: {"age_days": 2083, "days_push": 562, "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 1008, forks 228 (observed 2026-08-28T04:03:12.606354+00:00)

## What it is
The official baseline deep learning model for the GraspNet-1Billion benchmark, detecting dense 6-DoF grasp poses from point clouds of cluttered scenes. It includes training, inference, and evaluation code built on PyTorch with custom CUDA operators.

## Use cases
- detect grasp poses for objects from point clouds
- train a robot grasping detection model
- evaluate grasp detection on the GraspNet-1Billion benchmark
- run grasp inference on RGBD camera data
- reproduce CVPR 2020 grasping baseline results
- generate tolerance labels for grasp training

## When to choose
- you need a strong baseline for 6-DoF grasp detection research
- you want to benchmark against GraspNet-1Billion
- you're doing robotics manipulation research with point clouds

## When to avoid
- you need a production-ready grasping system for commercial use (license is CC BY-NC-SA)
- you want a plug-and-play robot control stack rather than a research model
- you lack a GPU or the ability to compile custom CUDA operators

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, computer-vision, robotics
- domain: robotics, computer-vision, deep-learning, autonomous-vehicles
- platform: python
- tags: grasp-detection, point-cloud, pytorch, benchmark, 6d-pose, rgb-d, research-code, non-commercial-license, linux, gpu

## Member repositories
- graspnet/graspnet-baseline (main) score 35

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:12.606354+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 2026-08-30T07:12:09.544087+00:00, confidence not recorded.
  - readme: https://github.com/graspnet/graspnet-baseline (fetched 2026-08-28T04:03:12.606354+00:00, sha bd847ad19c4c)
  - homepage: https://graspnet.net/ (fetched 2026-08-29T13:12:36.783132+00:00, sha 7a0a5b98f698)
  - site_page: https://graspnet.net/about.html (fetched 2026-08-29T13:12:36.792895+00:00, sha 756cf3d5a4af)
- Data as of 2026-08-30T08:39:29.467469+00:00.