ASPset-510 (Australian Sports Pose Dataset) is a large-scale video dataset for the training and evaluation of 3D human pose estimation models. It contains 17 different amateur subjects performing 30 sports-related actions each, for a total of 510 action clips.

This repository contains Python code for working with ASPset-510.

If you don't want to use these scripts and would prefer to directly download the data yourself, ASPset-510 is available on the Internet Archive at



$ conda env create -f environment.yml

GUI (Optional)

$ conda env update -f environment-gui.yml

PyTorch (Optional)

$ conda env update -f environment-torch.yml


Downloading the dataset downloads and extracts ASPset-510 data.

Example usage:

$ python src/aspset510/bin/ --data-dir=./data

Note that by default the original archive files will be downloaded and kept in the archives subdirectory of whichever path you set using --data-dir. To set a different path for the archives, use the --archive-dir option. To download the archives without extracting them, use the --skip-extraction option.

Browsing clips from the dataset provides a graphical user interface for browsing clips from ASPset-510.

Example usage:

$ python src/aspset510/bin/ --data-dir=./data


Acknowledgments and license

ASPset-510 is brought to you by La Trobe University and the Australian Institute of Sport. It is dedicated to the public domain under the CC0 1.0 license.

If you find this dataset useful for your own work, please cite the following paper:

  title={{ASPset}: An Outdoor Sports Pose Video Dataset With {3D} Keypoint Annotations},
  author={Nibali, Aiden and Millward, Joshua and He, Zhen and Morgan, Stuart},
  journal={Image and Vision Computing},