Future work

1. Hyperparameter optimization.
2. Experiment with different data augmentation strategies.
   • Apply random (fc, bw) bandpass filter.
   • Apply random power threshold cutoff.

Training

python train.py \
    --output_dir=./project_dir/final \
    --dataset_dir=./data/intermediate/ \
    --batch_size=32

The training scripts follows the model implementation provided in the paper (Learning Sleep Stages from Radio Signals: A Conditional Adversarial Architecture) with slight modification to the training optimization steps.

Predition

python pred.py \
    --input_dir=./project_dir/final