Attention-based_Atrous_CNN

Pytorch code for the paper 'Attention-based Atrous Convolutional Neural Networks: Visualisation and Understanding Perspectives of Acoustic Scenes', by Zhao Ren, Qiuqiang Kong, Jing Han, Mark Plumbley, Björn Schuller.

Data

DCASE 2018 Task 1 - Acoustic Scene Classification, containing:

subtask A: data from device A

subtask B: data from device A, B, and C

Preparation

channels:

• pytorch dependencies:
• matplotlib=2.2.2
• numpy=1.14.5
• h5py=2.8.0
• pytorch=0.4.0
• pip:
  • audioread==2.1.6
  • librosa==0.6.1
  • scikit-learn==0.19.1
  • soundfile==0.10.2

Run

sh runme.sh

In runme.sh, please run the following files for different tasks:

1. feature extraction: utils/features.py
2. training a model, and evaluation: main_pytorch.py

Cite

If the user referred the code, please cite our paper,

@InProceedings{ren2019attention,

Title = {{Attention-based atrous convolutional neural networks: Visualisation and understanding perspectives of acoustic scenes}},

Author = {Ren, Zhao and Kong, Qiuqiang and Han, Jing and Plumbley, Mark and Schuller, Bj"orn},

Booktitle = {Proc.\ ICASSP},

Year = {2019},

Address = {Brighton, UK},

Pages = {56--60}

}

Zhao Ren

Chair of Embedded Intelligence for Health Care and Wellbeing

University of Augsburg

07.08.2019