gitwtfhub

wtf is videopose3d?

facebookresearch/videopose3d — explained in plain English

Analysis updated 2026-07-03 · repo last pushed 2022-12-10

★ 4,036PythonAudience · developerComplexity · 3/5DormantLicenseSetup · moderate

TL;DR

Estimates 3D body joint positions from regular video by tracking how 2D joint locations change over time, useful for fitness apps, sports analysis, and motion capture without expensive studio equipment.

Mindmap

mindmap
  root((videopose3d))
    What it does
      3D pose estimation
      Joint tracking
      Motion analysis
    How it works
      2D keypoint input
      Temporal modeling
      3D lifting
    Use Cases
      Fitness form tracking
      Sports analytics
      Motion capture
    Tech Stack
      Python
      PyTorch
    Audience
      ML engineers
      Computer vision devs

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

Why would anyone build with this?

REASON 1

Build a fitness app feature that tracks and scores a user's exercise form using just a smartphone camera.

REASON 2

Analyze athlete technique in sports footage by extracting 3D body positions from game video.

REASON 3

Create a lightweight motion-capture pipeline for animation or special effects without studio hardware.

REASON 4

Run human pose estimation on your own custom video footage using the included pretrained models.

What's in the stack?

PythonPyTorch

How it stacks up

facebookresearch/videopose3dfacebookresearch/dlrmkarpathy/makemore
Stars4,0364,0484,010
LanguagePythonPythonPython
Last pushed2022-12-102026-01-122024-06-04
MaintenanceDormantQuietDormant
Setup difficultymoderatehardeasy
Complexity3/54/52/5
Audiencedeveloperdeveloperresearcher

Figures from each repo's GitHub metadata at analysis time.

How do you spin it up?

Difficulty · moderate Time to first run · 30min

Requires a separate 2D keypoint detector to generate inputs, training from scratch takes about a day on a high-end GPU.

Apache 2.0, free for any use including commercial, keep the license notice.

Yoink these prompts

Prompt 1
Using facebookresearch/videopose3d, write Python code to run 3D pose estimation on a local video file and visualize the skeleton.
Prompt 2
Show me how to download and use a pretrained VideoPose3D model to extract joint positions from a fitness workout video.
Prompt 3
How do I feed my own 2D keypoint detections into facebookresearch/videopose3d to get 3D pose output?
Prompt 4
What is the recommended pipeline for processing a custom video with facebookresearch/videopose3d from raw frames to 3D coordinates?

Frequently asked questions

wtf is videopose3d?

Estimates 3D body joint positions from regular video by tracking how 2D joint locations change over time, useful for fitness apps, sports analysis, and motion capture without expensive studio equipment.

What language is videopose3d written in?

Mainly Python. The stack also includes Python, PyTorch.

Is videopose3d actively maintained?

Dormant — no commits in 2+ years (last push 2022-12-10).

What license does videopose3d use?

Apache 2.0, free for any use including commercial, keep the license notice.

How hard is videopose3d to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is videopose3d for?

Mainly developer.

View the repo → Decode another repo

This repo across BitVibe Labs

Don't trust strangers blindly. Verify against the repo.

Other repos in this lane