/CAGE4HAR

Primary LanguagePython

CAGE

Contrastive Accelerometer-Gyroscope Embedding Model for Human Activity Recognition

Inyong Koo, Yeonju Park, Minki Jeong, Changick Kim

CAGE

This repository provides a pytorch implementation for the paper "Contrastive Accelerometer-Gyroscope Embedding Model for Human Activity Recognition (IEEE Sensors Journal, 2022)".

Data Preparation

  1. Download datasets (UCI-HAR, mHealth, MobiAct, PAMAP2) in data/

    data
    ├── UCI HAR Dataset
    ├── MHEALTHDATASET
    ├── PAMAP2_Dataset
    └── MobiAct_Dataset_v2.0

  • Although not introduced in paper, we also support Opportunity and USC-HAD datasets in our framework.
  1. Create processed data splits for all datasets with
python create_dataset.py

Training

To train Contrastive Accelerometer-Gyroscope Embedding (CAGE) model, run

python train_CAGE.py --dataset {Dataset}

We also provide baseline models such as DeepConvLSTM (Ordóñez et al., 2016), LSTM-CNN (Xia et al., 2020), and ConvAE (Haresamudram et al., 2019).

python train_baseline.py --dataset {Dataset} --model {Model}