As a first step,
-
extract the Compare dataset into the 'dataset' folder and rename the folder as 'Compare-Data'. Then extract the CCS data in the 'Compare-Data' folder into that folder.
-
Extract the Musan dataset into the 'dataset' folder and rename the folder as 'Musan-Data'.
The next step is as follows
-
Run the code '1_Download_Dataset.ipynb', this code will download the Coswara and Coughvid datasets, then extract them into the 'dataset' folder
-
Run the code '2_Create_CSV_Files.ipynb', this code will create csv data regarding the Coswara, Coughvid, and Compare datasets and will be saved into the 'csv_files' folder
-
Run the code '3_Normalized.ipynb', this code will normalize the dataset to be used, and the normalization results will be in each folder in the dataset as in 'dataset/Coswara-Data/wav_normalized'. This code will also merge the 3 normalized datasets into one folder named 'dataset/Merge-Data' and also create CSV data related to the merged data in the 'csv_files/normalized_data' folder. In addition, CSV data will be created regarding train data, eval data, and test data into the 'csv_files/experiment_data' folder.
-
Run the code '4_Training.ipynb', this code will train the combined dataset using 5 random seeds. In this code, seeds 9,30,41,42, and 46 are used because they have high yields. The training results will be saved in the 'results' folder.
-
Run the code '5_Testing.ipynb', this code will test the model that has been trained using the test data. This test is carried out using 5 seeds that have been trained previously.