Baseline-MATLAB-DCASE

This repo contains an unofficial MATLAB implementation of DCASE2021 Task 1A baseline code, which is part of the DCASE challenge.

Unofficial Baseline System for DCASE2021 Task1A Using MATLAB®

This repo contains an unofficial MATLAB implementation of DCASE2021 Task 1A baseline code, which is part of the DCASE 2021 challenge.

Copyright 2021 The MathWorks, Inc.

MathWorks Products (https://www.mathworks.com)

Requires MATLAB release R2021a or newer. To train the baseline system, the following toolboxes are required:

• Audio Toolbox™
• Deep Learning Toolbox™

To accelerate training, the following toolbox is recommended:

• Parallel Computing Toolbox™

To quantize a network, the following package is required:

• Deep Learning Toolbox Model Quantization Library

To deploy a quantized network to CUDA code, the following toolbox is required:

• GPU Coder™

See Quantization Workflow Prerequisites for a list of required products depending on your target.

Known Differences with Official Baseline System

This unofficial baseline has the following known differences with the official baseline. There may be additional differences.

1. The mini-batch size was increased from 16 to 256.
2. A piecewise learn rate schedule was added with a drop period of 100 epochs. The max number of epochs was reduced from 200 to 120.
3. This example uses and evaluates the final state of the network, after all epochs are complete. The official baseline uses the best peforming model over all of the epochs.
4. This example only trains the network and evalutes the system once. The official baseline trains and evaluates 10 times to provide additional statistical analysis.
5. This example uses int8 quantization instead of the half-precision quantization in the official baseline.

Getting Started

To run this baseline, add Unofficial_DCASE2021_Task1A_Baseline_Using_MATLAB.mlx and classifyAcousticScene.m to your current folder in MATLAB and then run Unofficial_DCASE2021_Task1A_Baseline_Using_MATLAB.mlx. The example loads and examines the data, defines and trains a model, quantizes the model, and evaluates the quantized model.

You can view a PDF of the executed example in the file Unofficial_DCASE2021_Task1A_Baseline_Using_MATLAB.pdf.

License

The license is available in the License.txt file in this repository.

References

[1] Irene Martin-Morato, Toni Heittola, Annamaria Mesaros, and Tuomas Virtanen. Low-Complexity Acoustic Scene Classification for Multi-Device Audio: Analysis of DCASE 2021 Challenge Systems. 2021. URL: https://arxiv.org/abs/2105.13734.