Action Recognition for Privacy-Preserving Ambient Assisted Living (AIiH2024)

This is the official repository for Action Recognition for Privacy-Preserving Ambient Assisted Living, our paper published at the International Conference on AI in Healthcare (AIiH2024)

The paper can be found here

Prerequisites

Python >= 3.6

PyTorch >= 1.1.0

PyYAML, tqdm

Run pip install -e torchlight `

Data Preparation

Download datasets.

There are two datasets to download:

NTU RGB+D 60 Skeleton

NW-UCLA

1. NTU RGB+D 60: Download the Skeleton dataset here
2. NW-UCLA: Download the dataset here

Data Processing

Directory Structure

Put downloaded data into the following directory structure.

- data/
  - NW-UCLA/
    - all_sqe
      ...
  - ntu/
    - nturgbd_raw/
	  - nturgb+d_skeletons
            ...

Generating Data

NW-UCLA dataset

Move folder all_sqe to ./data/NW-UCLA

NTU RGB+D 60 dataset

First, extract all skeleton files to ./data/ntu/nturgbd_raw
 cd ./data/ntu
 # Get the skeleton of each performer
 python get_raw_skes_data.py
 # Remove the bad skeleton 
 python get_raw_denoised_data.py
 # Transform the skeleton to the centre of the first frame
 python seq_transformation.py

Training

NTU RGB+D 60 dataset:

For cross-view, run python main.py --device 0 1 --config ./config/nturgbd-cross-view/default.yaml

For cross-subject, run python main.py --device 0 1 --config ./config/nturgbd-cross-subject/default.yaml

NW-UCLA dataset:

Run python main.py --device 0 1 --config ./config/ucla/nw-ucla.yaml

Testing

NTU RGB+D 60 dataset:

For cross-view, run python main.py --device 0 1 --config ./config/nturgbd-cross-view/default.yaml --phase test --weights path_to_model_weight

For cross-subject, run python main.py --device 0 1 --config ./config/nturgbd-cross-subject/default.yaml --phase test --weights path_to_model_weight

NW-UCLA dataset:

Run python main.py --device 0 1 --config ./config/ucla/nw-ucla.yaml --phase test --weights path_to_model_weigh

Real-time action recognition in a real-world environment

Structured environment

An environment with no occlusions

Unstructured environment

An environment with occlusions and noise

Citation

@InProceedings{10.1007/978-3-031-67285-9_15,
author="Zakka, Vincent Gbouna
and Dai, Zhuangzhuang
and Manso, Luis J.",
editor="Xie, Xianghua
and Styles, Iain
and Powathil, Gibin
and Ceccarelli, Marco",
title="Action Recognition for Privacy-Preserving Ambient Assisted Living",
booktitle="Artificial Intelligence in Healthcare",
year="2024",
publisher="Springer Nature Switzerland",
address="Cham",
pages="203--217",
isbn="978-3-031-67285-9"
}