openpack-torch 0.3.0

PyTorch extention to work around with OpenPack dataset

openpack-torch

PyTorch utilities to work around with OpenPack Dataset.

Setup

You can install via pip with the following command.

# Pip
pip install openpack-torch

# Poetry
poetry add  openpack-torch

Docs

• Dataset Page
• API Docs
• PyPI - openpack-torch

Examples

Operation Recognition (Semantic Segmentation)

IMU

• Acceleration
  • U-Net
  • DeepConvLSTM

Vision

• Keypoints
  • ST-GCN

Scores of Baseline Moodel (Preliminary Experiments)

Split: Pilot Challenge

Model	F1 (Test Set)	F1 (Submission Set)	Date	Code
UNet	0.3451	0.3747	2022-06-28	main.py
DeepConvLSTM	0.7081	0.7695	2022-06-28	main.py
DeepConvLSTM + Self-Attn	0.8161	0.8409	2022-06-28	main.py
ST-GCN	-	-

NOTE: F1 = F1-measure (macro average)

LICENCE

This software (openpack-torch) is distributed under MIT Licence. For the license of "OpenPack Dataset", please check this site (https://open-pack.github.io/).