# ViTAE-Transformer/ViTPose

The official repo for [NeurIPS'22] "ViTPose: Simple Vision Transformer Baselines for Human Pose Estimation" and [TPAMI'23] "ViTPose++: Vision Transformer for Generic Body Pose Estimation"

Repository: https://github.com/ViTAE-Transformer/ViTPose
Canonical: https://ross.abutalabs.com/products/vitpose
Language: Python
License: Apache-2.0
License Family: permissive
Topics: deep-learning, distillation, mae, pose-estimation, pytorch, self-supervised-learning, vision-transformer
Last push: 2025-12-25T16:25:15+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 59, release rhythm 35, longevity 100
- inputs: {"age_days": 1590, "days_push": 251, "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 2138, forks 268 (observed 2026-08-28T04:06:18.203741+00:00)

## What it is
Official PyTorch implementation of ViTPose and ViTPose++, Vision Transformer models for human and generic body pose estimation from NeurIPS 2022 and TPAMI 2023 papers. It provides pretrained checkpoints, training/evaluation configs, and achieves state-of-the-art results such as 81.1 AP on MS COCO keypoints.

## Use cases
- estimate human body keypoints from images
- run pose estimation on video
- fine-tune a vision transformer for pose estimation
- benchmark pose estimation models on MS COCO
- use MAE-pretrained backbones for keypoint detection
- distill large pose models into smaller ones

## When to choose
- you need high-accuracy 2D human pose estimation in PyTorch
- you want a simple vision transformer baseline for keypoints
- you need pretrained checkpoints for COCO-style pose tasks

## When to avoid
- you need 3D pose estimation or full body mesh recovery
- you want a plug-and-play production API rather than research code
- you work outside PyTorch or lack GPU resources

## Facets
- artifact type: library
- maturity: stable
- function: machine-learning, deep-learning, computer-vision
- domain: computer-vision, deep-learning, machine-learning
- platform: python
- tags: pose-estimation, vision-transformer, pytorch, human-pose, keypoint-detection, mae-pretraining, knowledge-distillation, research-code, gpu

## Member repositories
- ViTAE-Transformer/ViTPose (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:06:18.203741+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 2026-08-30T02:51:35.653724+00:00, confidence not recorded.
  - readme: https://github.com/ViTAE-Transformer/ViTPose (fetched 2026-08-28T04:06:18.203741+00:00, sha a41d5fa36897)
- Data as of 2026-08-30T08:39:29.467469+00:00.