Link: https://dl.acm.org/doi/abs/10.1145/3460426.3463619
This repo inherits the implementation of the MGC model proposed in a recent paper - MS-SincResNet - https://arxiv.org/abs/2109.08910. We are able to acheive a better performance of 91.50% than the original paper. We propose various augmentation strategies and combination of center and label smoothing losses to achieve better test accuracy. Additionally, we have also extended our model on Extended Ballroom Dataset and FMA dataset and are having competitive results with other SOTA methods.
- Make sure you have python 3.6 and above
- Run below command to install dependencies
pip install -r requirements.txt
Running on single GPU:
python Main.py
Running on two GPUs:
bash run.sh
- GTZAN will be auto downloaded
- To get Extended Ballroom, please first run
python preprocess/script/getEBallroom.py
- To get FMA_small please run
cd _data/FMA_small wget https://os.unil.cloud.switch.ch/fma/fma_small.zip wget https://os.unil.cloud.switch.ch/fma/fma_metadata.zip python make_label_FMA_small.py
Official implementation of the original paper: https://github.com/PeiChunChang/MS-SincResNet
Multiple GPUs currently unavailable.