时空索引>> Audio-Classification-Spectrogram>> 返回
项目作者： favia96

项目描述 ：
Audio classification using spectrograms and transfer learning
高级语言： Python
项目地址: git://github.com/favia96/Audio-Classification-Spectrogram.git
创建时间： 2020-09-02T12:02:44Z
项目社区：https://github.com/favia96/Audio-Classification-Spectrogram
开源协议：MIT License
下载


Audio classification using spectrograms and transfer learning

Project 1 - Research Methodology and Scientific Writing (EQ2425) - Federico Favia & Mayank Gulati, December 2019, KTH, Stockholm
This project in Python developed within the course of esearch Methodology and Scientific Writing at KTH is a pratical and theoretical research about audio classification on musical instruments of NSynth audio dataset using Convolution Neural Network (CNN) technique by converting .wav audio file to spectrogram images. In this project, the open source Matlab library VLFeat for SIFT features is used. We exploit on the existing state-of-the-art infrastructures of image classification used in ImageNet challenge namely ResNet18, AlexNet, GoogLeNet with the help of transfer learning by adding custom layers in the end of these deep neural networks. See the report for more information.
Below you can see the spectrograms used:
Spectrogram