Model pre-trained on Deep soli dataset and fine-tuned on self-collected dataset recorded using BGT60TR13C 60 GHz FMCW radar. It consists of only 350 classes over 7 classes recorded by only one person (the dataself is not included). Therefore the model is not production-ready. The model itself is a combination of CNN for spacial features extraction and LSTM for temporal features extraction.
4uf04eG/FMCW-gesture-recognition
CNN + LSTM for hand gesture recognition with BGT60TR13C FMCW radar
Jupyter Notebook