This is the official repository for Action Recognition in Real-World Ambient Assisted Living Environment, our paper submitted to the Journal of Big Data Mining and Analytics

Use the following guide to set up the training environment.

1. Create a conda environment
2. Install pytorch
3. Then install the dependencies using this command python3 -m pip install -r requirements.txt

Alternatively, you can use our environment.yml file to create an environment with all dependencies. Note that this environment was created on Ubuntu 22.04.4 LTS with cuda version 11.4.

Create the environment using the following command

conda env create -f environment.yml

Then, activate it using the following command

conda activate GCN

1. NTU RGB+D 60: Download the Skeleton dataset here
2. NW-UCLA: Download the dataset here
3. SHREC'17: Download the dataset here

Put the downloaded data into the following directory structure.

- data/
  - NW-UCLA/
    - all_sqe
      ...
  - ntu/
    - nturgbd_raw/
	  - nturgb+d_skeletons
            ...
  - shrec/
    - shrec17_dataset/
	  - HandGestureDataset_SHREC2017/
	    - gesture_1
	      ...

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

SHREC'17 dataset

First, extract the downloaded dataset to /data/shrec/shrec17_dataset

Then, run python gen_traindataset.py and python gen_testdataset.py to prepare the dataset

Note: The --device 0 1 argument in the training and testing command specifies the GPU indices to be used.

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

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

Run python main.py --device 0 --config ./config/shrec17/shrec17.yaml

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

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

Run python main.py --device 0 1 --config ./config/shrec17/shrec17.yaml --phase test --weights path_to_model_weigh