Music Emotion Recognition Research

This is a research about music emotion recognition conducted by Alex, Quan, and Rick.

• Result folders: https://drive.google.com/drive/folders/1VDdX_ydLfx_SG5bgtZyxEzB4ob_eDv0_?usp=sharing

Folder structure of the project

• /configs/: contains global variables, setup, environment variables, training variables, and every other configurations to be plugged into the tool.
• /data/: contain music emotion datasets
• /docs/: contains resource, figures, and articles. Some of them are referred from the backlog.md file.
• /mains/: main files to be run
• /mer/: music emotion recognition code base library.
  • /mer/mer/: music emotion recognition code base library
    • /mer/dataloader/: contains files to preprocess data
    • /mer/model/: contains model builder files
  • /mer/setup.py: mer setup file
• /models/: contains trained models and checkpoint
• /mss/: music source separation code base library
  • /mss/mss/: music source separation code base library
    • /mss/base/: contains base file (model and trainer)
    • /mss/dataloader/: contains dataloader file
    • /mss/model/: contains model builder files
    • /mss/trainer/: contains trainer files
    • /mss/utils/: contain utility files
  • /mss/setup.py: mss setup file
• /notebooks/: contains notebook files for easy experiments.
• /old/: old experiment code base files, used for reference.

How to test and experiment

(to be implemented)

Get the dataset to be generated

1. python ./old/wav_converter.py "./data/PMEmo/PMEmo2019/PMEmo2019/chorus" "./data/PMEmo/PMEmo2019/PMEmo2019/chorus_wav" ffmpeg

2. python Predict.py with cfg.full_multi_instrument model_path="./checkpoints/full_multi_instrument/full_multi_instrument-134067" input_path="../data/PMEmo/PMEmo2019/PMEmo2019/chorus/" output_path="../data/PMEmo/PMEmo2019/PMEmo2019/separation/"

3. python ./mains/convert_16_bit_depth.py ./data/PMEmo/PMEmo2019/PMEmo2019/separation ./data/PMEmo/PMEmo2019/PMEmo2019/separation_16