/ Machine Learning

Multi-Stage Temporal Convolutional Network for Action Segmentation

Multi-Stage Temporal Convolutional Network for Action Segmentation

MS-TCN

This repository provides a PyTorch implementation of the paper MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation.

Tested with:

  • PyTorch 0.4.1
  • Python 2.7.12

Qualitative Results:

Training:

  • Download the data folder, which contains the features and the ground truth labels. (~30GB)
  • Extract it so that you have the data folder in the same directory as main.py.
  • To train the model run python main.py --action=train --dataset=DS --split=SP where DS is breakfast, 50salads or gtea, and SP is the split number (1-5) for 50salads and (1-4) for the other datasets.

Prediction:

Run python main.py --action=predict --dataset=DS --split=SP.

Evaluation:

Run python eval.py --dataset=DS --split=SP.

Citation:

If you use the code, please cite

Y. Abu Farha and J. Gall.
MS-TCN: Multi-Stage Temporal Convolutional Network for Action Segmentation.
In IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2019

GitHub