Pytorch implementation of Sample-level Deep Convolutional Neural Networks for Music Auto-tagging Using Raw Waveforms
- Used tag annotations and audio data
3^9 model with input sample size 59049
3 : stride length of the first conv layer (along with filter size 3, it reduces input dimension to 19683)
9 : 9 hidden conv layers
- Data processing
- audio (to read audio signal from mp3s and save as npy) :
python process_audio.py - annotation (process redundant tags and select top N=50 tags):
python process_annotations.py- this will create and save train/valid/test annotation files
- audio (to read audio signal from mp3s and save as npy) :
- Training
python main.py --device_num 0
- Testing
- predict tags for given songs
python evaluate.py --device_num 0
python eval_tags.py --device_num 0 --mp3_file "path/to/mp3file/to/predict.mp3"