This is the official repository for Hierarchical Temporal Convolution Network: Towards Privacy-Centric Activity Recognition, our paper published at the International Conference on Ubiquitous Computing and Ambient Intelligence (UCAmI)
The paper can be found here
Python >= 3.6
PyTorch >= 1.1.0
PyYAML, tqdm `
Download the JHMDB and SHREC datasets using the links below:
JHMDB raw data download link: http://jhmdb.is.tue.mpg.de/challenge/JHMDB/datasets
SHREC raw data download link: http://www-rech.telecom-lille.fr/shrec2017-hand/
Use the preprocessing code in the data processing folder to process the data and put them in the data folder.
For JHMDB, run python train.py --batch-size 512 --epochs 600 --dataset 0 --lr 0.001 | tee train.log
For SHREC coarse, run python train.py --batch-size 512 --epochs 600 --dataset 1 --lr 0.001 | tee train.log
For SHREC fine, run python train.py --batch-size 512 --epochs 600 dataset 2 --lr 0.001 | tee train.log
To test the trained model, bring the saved model to the main directory and pass its name as an arg for the model-path or simply pass the path to where the model was saved
For JHMDB, run python test.py --model-path model.pt --dataset 0
For SHREC coarse, run python test.py --model-path model.pt --dataset 1
For SHREC fine, run python test.py --model-path model.pt --dataset 2
To force the model to be loaded with CPU run python test.py --model-path model.pt --dataset 0 --no-cuda
@InProceedings{HT-ConvNet,
author="Zakka, Vincent Gbouna
and Dai, Zhuangzhuang
and Manso, Luis J.",
editor="Bravo, Jos{\'e}
and Nugent, Chris
and Cleland, Ian",
title="Hierarchical Temporal Convolution Network: Towards Privacy-Centric Activity Recognition",
booktitle="Proceedings of the International Conference on Ubiquitous Computing and Ambient Intelligence (UCAmI 2024)",
year="2024",
publisher="Springer Nature Switzerland",
address="Cham",
pages="344--356",
isbn="978-3-031-77571-0"
}