This project implements a deep learning model for classifying objects such as humans, cats, and other creatures using WiFi Channel State Information (CSI) signals. The model is trained on the Animal Crossing WiFi CSI Dataset and utilizes an LSTM-based architecture to process sequential signal data. The model predicts the object class directly from the WiFi signal data.
The dataset used in this project is the Animal Crossing WiFi CSI Dataset, which contains WiFi CSI signals labeled according to the object/creature present, such as humans, cats, and other animals. The dataset captures continuous WiFi signals and represents them as time-series data across multiple subcarriers.
- Dataset Link: Animal Crossing WiFi CSI Dataset on Zenodo
- Access: The dataset is provided as
TRAIN.parquetandTEST.parquetfiles for training and evaluation.
- Continuous CSI signals are reshaped into a 3D tensor
(num_samples, TIME_FRAMES, N_SUBCARRIERS)suitable for LSTM input. - Labels are encoded as integers corresponding to each object class.
- An LSTM-based neural network is used to capture temporal dependencies in WiFi signals.
- The network consists of multiple LSTM layers followed by a fully connected layer to output class predictions.
- The model is trained using cross-entropy loss and optimized with the Adam optimizer.
- GPU acceleration is used if available.
- Python 3.x
- PyTorch
- NumPy
- pandas
- The model successfully classifies objects from WiFi CSI signals.
- Model performance is evaluated by comparing predicted object classes with ground truth labels on the test dataset.
This project is licensed under the MIT License. See the LICENSE file for more details.