This is the official repository for Multi-Head Adaptive Graph Convolution Network for Sparse Point Cloud-Based Human Activity Recognition

Create conda environment with python 3.8

Install cuda toolkit using the command below from this link https://anaconda.org/nvidia/cuda-toolkit
conda install nvidia/label/cuda-11.8.0::cuda-toolkit

Then, install the following:
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=11.8 -c pytorch -c nvidia
conda install conda-forge::tqdm
conda install conda-forge::pytorch_geometric

Download the MiliPoint dataset from their Google drive or from the GitHub repo. Unzip the downloaded data and put the contents in data/raw/ according to the file structure below.

In the Milipoint folder, according to the file structure below, make a directory data/processed/mmr_action, where the processed data will be stored.

The data.py script in the MiliPoint directory prepares the dataset for training as specified by the data providers.

MiliPoint
└─data
  └─raw
    ├─0.pkl
    ├─1.pkl
    ├─...
  └─processed
    └─mmr_action

Download the MMActivity dataset from their GitHub repo

The data consist of two folders: train and test. Each of these folders further contains subfolders corresponding to the respective activity classes.

Then, run the process.py script to prepare the data. This will generate pickle files for each action class in the train and test folders. Copy the generated pickle files to the corresponding train and test folders in the data/raw directory, following the file structure below.

In the MMActivity folder, according to the file structure below, make a directory data/processed/mmr_action, where the processed data will be stored.

The data.py script in the MMActivity directory prepares the dataset for training as specified by the data providers.

MMActivity
└─data
  └─raw
    └─train
      ├─0.pkl
      ├─1.pkl
      ├─...
    └─test
      ├─0.pkl
      ├─1.pkl
      ├─...
  └─processed
    └─mmr_action

First, go into the MiliPoint directory, then run python train.py --use_sgd to train and python train.py --eval to test the trained model.

First, go into the MMActivity directory, then run python train.py --use_sgd to train and python train.py --eval to test the trained model.