AMC Parser

A lightweight project to parse and visualize asf/amc files from CMU MoCap dataset using Python 3. GitHub link: AMCParser.

The main purpose of this project is to help understanding how asf/amc file works, as well as how to use them. Therefore, simplicity and readability are stressed, while robustness is ignored.

Demo

Demo using PyGame and PyOpenGL:

3D Demo

Demo using Matplotlib:

Static Demo

Usage

There’s a simple example in the source code as follows:

if __name__ == '__main__':
  asf_path = './data/01/01.asf'
  amc_path = './data/01/01_01.amc'
  joints = parse_asf(asf_path)
  motions = parse_amc(amc_path)
  frame_idx = 180
  joints['root'].set_motion(motions[frame_idx])
  joints['root'].draw()

And another exampls in 3Dviewer.py:

  asf_path = './data/01/01.asf'
  amc_path = './data/01/01_01.amc'
  joints = parse_asf(asf_path)
  motions = parse_amc(amc_path)
  v = Viwer(joints, motions)
  v.run()

The data can be found from CMU MoCap dataset.

Parser

The asf/amc parsers are straightforward and easy to understand. The parsers are fully tested on the CMU MoCap dataset, but I don’t expect it can work on other datasets without any modification. However, it won’t be hard to extend it for more complicating asf/amc files.

Visualization

Matplotlib is used to draw joints and bones in 3D statically; PyGame and PyOpenGL are used to draw motion sequence.

In 3DViewer, we support:

  • WASD to move around.
  • QE to zoom in/out.
  • ↑ ↓ ← → to rotate.
  • LEFT MOUSE BUTTON to drag.
  • RETURN to reset camera view.
  • SPACE to start/pause.
  • , and . to forward and rewind.

NOTE that my implementation of changing view is inefficient (but fluent enough) with some small bugs, but I don’t have time to improve it. Pull request is welcomed.

Dependencies

  • numpy
  • transforms3d
  • matplotlib
  • pygame
  • pyopengl

All the dependencies are available via pip install.