WiFi-CSI-Sensing-Behavior-Recognition

Train CNN, MLP, ResNet18, RNN, LSTM on CSI data to classify 7 behaviors.walk, stand, sit, run, fall, bow, null

Requirements

python==3.11.3
torch==2.0.0+cu117
pandas==2.0.2

Dataset

CSI data of 30*900 every sample,there exist 600 samples.You can change to your own dataset by modifing dataset_process.py and main.WifiDataset().
Dataset used here is not offered!

Run

python main.py --model MLP --epoch 50 --gpu cuda:0 --batch_size 128
For --model, you can choose from MLP, ResNet18, RNN, LSTM, CNN_GRU.

If you want to train model from scratch, add --fromScratch parameter, for example:
python main.py --model MLP --epoch 50 --gpu cuda:0 --batch_size 128 --fromScratch
Or using follow to train a model from checkpoint:
python main.py --model MLP --epoch 50 --gpu cuda:0 --batch_size 128 --checkpoint ./checkpoint/checkpoint.pth

Inference

Only inference using saved checkpoint:
python main.py --model MLP --checkpoint ./checkpoint/checkpoint.pth