Spaces:

kehong
/

MoCapAnythingV2

Running on Zero

App Files Files Community

MoCapAnythingV2 / README.md

kehong

Fix subtitle: drop 'from Monocular Videos' (that is the V1 title)

6a0e103 verified 27 days ago

preview code

Raw History Blame Contribute Delete

1.14 kB

A newer version of the Gradio SDK is available: 6.29.0

Upgrade

metadata

title: MoCapAnything V2
emoji: 🐾
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 6.20.0
app_file: app.py
short_description: Motion Capture & Retargeting for Arbitrary Skeletons
license: mit
models:
  - kehong/MoCapAnythingV2-weights
datasets:
  - kehong/MoCapAnythingV2-data-sample
tags:
  - motion-capture
  - keypoint-detection
  - retargeting
  - animation
  - animals

🐾 MoCapAnything V2

End-to-End Motion Capture for Arbitrary Skeletons — pick a video, pick a target species or object (it can differ from the input → retargeting), and get the predicted 3D pose skeleton plus BVH/.npy downloads.

📄 Paper: MoCapAnything V2 on HF Papers
💻 Code: github.com/phongdaot/MocapAnything
🏋️ Weights: kehong/MoCapAnythingV2-weights

This Space shows the skeleton output; the local demo in the repo additionally offers the Blender 3D mesh render and the 💃 Dance Anything tab (dance video + music → an animal performs the dance).