This is the repository to store the codes and materials in the KDD group project for Group (No. 2)
- Project topic: classification
- Group number: N0.2
Topic: classification.
Title: COVID-19 Classification Based on Cough Sound
Chung-chi,
Jiabao,
Zongchao,
Jihong,
Yubo,
Lingyun
The data we are using is collected from Coswara.
The models applied in this project are as follows:
- VAE
- GRU
- LSTM
- VGGish
- transformer-AST
The results are shown in the final paper.
how to compile and execute the description of each source file an example to show how to run the program the operating system you tested your program (e.g., linux and Windows) anything you want to include
Linux
python clean2_readTheFile.py # to clean the audio files with less audio files
python clean2_readFileOverviewCsv.py # to clean the audio files with damaged audio files
python mfcc.py # to extract the MFCC features
Linux
python load_npy.py # to use the saved data
python main.py # to run the VAE model
Environment:
Python 3.6, torch, torchnet, pandas, sklearn, tqdm
python main.py --data_dir={data_dir} --label={label}
--data_dir: Dir that store the MFCC preprocessed data, 'distribution/' by defult
--label: The csv file that store the preprocessed labels and other informations, 'combined_data_clean2.csv' by defult
After running the code, the processed dataset which used for input of the GRU model would be stored in the dataset folder.
Environment: Python 3.7, torch, numpy, sklearn
python main.py --base_dir /data/lingyun/coswara-data/ --hidden_dim 64 --num_layers 4 --dropout 0 --optimizer adam --lr 0.001
--base_dir: the base directory of the coswara dataset
More examples can be found in run.sh.
References are listed here.