AIRHAR (RadMamba)

AIRHAR is a learning framework built in PyTorch for radar-based human activity recognition. Developed by the Lab of Efficient Machine Intelligence @ Delft University of Technology, AIRHAR aims to bridge the gap between machine learning and signal processing of radar system.

The framework provides a comprehensive solution for training neural network models to classify non-continuous and continuous human acivities. By leveraging state-of-the-art deep learning techniques, AIRHAR enables researchers and engineers to develop more energy-efficient wireless communication systems.

We invite you to contribute to this project by providing your own backbone neural networks, pre-trained models, or measured radar-based human activities/gesture datasets.

Our latest work: RadMamba: Efficient Human Activity Recognition through Radar-based Micro-Doppler-Oriented Mamba State-Space Model

Project Structure

.
└── backbone        # Configuration files for classifiers
└── datasets        # datasets preparation files for open source datasets in RadMamba work
└── log             # Experimental log data (automatically generated)
└── modules         # Major functional modules
└── save            # Saved models
└── steps           # Implementation steps (classification .. to be continued)
└── utils           # Utility libraries
└── argument.py     # Command-line arguments and configuration
└── main.py         # Main entry point
└── model.py        # Top-level neural network models
└── project.py      # Core class for hyperparameter management and utilities

Environment Setup

This project has been tested with PyTorch 2.5.1 and Ubuntu 24.04 LTS.

Setting Up Your Environment

We recommend using Miniconda for environment management:

# Install Miniconda (Linux)
wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.sh

# For MacOS, use:
# wget https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh

# Create a Python environment with required packages
conda create -n AIRHAR python=3.13 numpy scipy pandas matplotlib tqdm adabound einops h5py scikit-learn cv2 
conda activate AIRHAR

Installing PyTorch

For Linux or Windows systems:

• With CPU only:

  pip3 install torch torchvision torchaudio

• With NVIDIA GPU (CUDA 12.6):

  pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu126

  Note: Ensure you have the latest NVIDIA GPU drivers installed to support CUDA 12.6

For macOS systems:

pip3 install torch torchvision torchaudio

Reproducing Published Results

Data Downloads

Please follow the instructions fro

Non-continuous CW Dataset (DIAT)

https://ieee-dataport.org/documents/diat-μradhar-radar-micro-doppler-signature-dataset-human-suspicious-activity-recognition

Non-continuous FMCW Dataset (CI4R)

https://github.com/ci4r/CI4R-Activity-Recognition-datasets

Continuous FMCW Dataset (UoG20)

UoG2020 dataset from Glasgow is being prepared for integration into the existing collection at https://researchdata.gla.ac.uk/848/.

Data Preparation

To reproduce the datasets preprocessing shown in RadMamba, please change the path in data_prepare.sh to your data path:

bash data_prepare.sh

This script prepares three datasets: non-continuous CW dataset (DIAT), non-continuous FMCW dataset (CI4R), and continuous FMCW dataset (UoG20)

Classification

To reproduce the classification results in Figure 4(a1) (a2) (a3), and Table IV:

bash run_experiments_DIAT.sh
bash run_experiments_CI4R.sh
bash run_experiments_UoG20.sh

These scripts train various classifier models across model sizes.

Authors & Citation

If you find this repository helpful, please cite our work:

• RadMamba: Efficient Human Activity Recognition through Radar-based Micro-Doppler-Oriented Mamba State-Space Model

@misc{wu2025radmambaefficienthumanactivity,
      title={RadMamba: Efficient Human Activity Recognition through Radar-based Micro-Doppler-Oriented Mamba State-Space Model}, 
      author={Yizhuo Wu and Francesco Fioranelli and Chang Gao},
      year={2025},
      eprint={2504.12039},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2504.12039}, 
}

Acknowledgment

This work was partially supported by the European Research Executive Agency (REA) under the Marie Skłodowska-Curie Actions (MSCA) Postdoctoral Fellowship program, Grant No. 101107534 (AIRHAR).

Contributors

• Chang Gao - Project Lead
• Yizhuo Wu - Core Developer