PeiChunChang / MS-SincResNet-1

A PyTorch implmentation of MS-SincResNet [ICMR 2021] "MS-SincResNet: Joint learning of 1D and 2D kernels using multi-scale SincNet and ResNet for music genre classification"

MS-SincResNet

Link: https://dl.acm.org/doi/abs/10.1145/3460426.3463619

About

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.

Setup

1. Make sure you have python 3.6 and above
2. 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

Currently support GTZAN and Extended Ballroom and FMA-small.

• 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

Link

Official implementation of the original paper: https://github.com/PeiChunChang/MS-SincResNet

Known issue

Multiple GPUs currently unavailable.