https://github.com/bikcrum/SongMultimodelAnalysis/blob/master/Report.pdf
- Download scripts from here https://drive.google.com/drive/folders/1G7fdJRXJj9w0n6hogd01xn2Wrj6mgqAd?usp=sharing
| Order | Script | Decription |
|---|---|---|
| 1. | create_empty_dataset_folders.sh | This creates empty directory where datasets will be downloaded. You must change working directory in this file but you should not change the following dataset path. |
| 2. | download_mp3.sh | Your mp3 songs will be downloaded |
| 3. | converter_wav.sh | Converts all your mp3 songs to wav format |
-
| feature_extraction.sh | Converts all wav song to melspectrogram and clean lyrics and saves in cache directory. [OPTIONS: al, a or l, where a=audio, l=lyrics] 5.| train.sh | Trains your model. You can view your tensorboard at http://localhost:6006
-
You must follow the order of execution as given above. For example.
converter_wav.shcannot be run beforedownload_mp3.sh
- Melspectrogram is converted from wav file in configuration specified in utils.py file.