Repetition code of the model for the paper "EEG-Based Objective Measurement of Temporal Profiles of Emotion Intensity" in pytorch
In this work, we are investigating emotion intensity profile based on EEG signals.
install pip3 install -r requirements.txt
argparse
torch==1.8.0
h5py==3.2.0
numpy==1.20.1
scikit-learn==0.24.1
scipy==1.6.2
python main.py
| Parameter | Default | Description |
|---|---|---|
| --data-path | DEAP raw file path | |
| --start-subjects | 0 | the starting subject number (from 0 to 31) |
| --subjects | 32 | the number of training subjects |
| --label-type | 'A' | emotion label ('A' : arousal 'V':valence) |
| --round | 5 | baseline training repeated round |
| --weight_decays | 1e-2 | optimizer weight decays |
| --save-path | './save' | the trained model saving file path |
| -testlist | False | if test list existed ( True: exist; False: split train/test trials) |
| --testlist_file | './save/arousal/' | test list file path |
| --max-epoch | 150 | training epochs for baseline and new model training |
| --ft-epoch | 50 | training epochs for fine tune training |
| --threshold | 20 | the percentage of selecting sub-dataset size |
| -retrain-type | 'ft' | retrain method ('ft': fine tune, 'nm': new model) |
| --selection | 'time' | selection method ('time': time interval method; 'score': score method) |
| --batch-size | 128 | batch size for training dataset |
.
├── github # Score code files (alternatively `dist`)
│ ├── main.py # Main file to execute research
│ ├── SCCNet.py # Model architecture
│ ├── channel_baseline.txt # Channel order for prepare data
│ ├── func.py # Functional tool use in research
│ ├── generate_profile.py # Code for generating emotion intensity value for trials
│ ├── preprocess.py # Prepare raw data in DEAP dataset
│ ├── requirements.txt # Requirement detail
│ ├── score.py # Score selection and retraining process
│ ├── time_interval.py # Time-interval selection and retraining process
│ └── training.py # Baseline model training process
└── README.md