Porting some of my old convnet code into Keras - a work in progress.
It's still a very rough skeleton, so use at your own risk, but due to numerous inquiries I decided to publish it sooner rather than later. I will try to implement the missing bits soon.
Current features:
examples/esc_convnet_paper.pyreimplements the Environmental Sound Classification with Convolutional Neural Networks paperechonet/datasets/esc_original.pyprovides a loading wrapper around the ESC-50 dataset
For the time being some important requirements:
- Python 3.5
- ESC-50 WAV version: https://github.com/karoldvl/ESC-50/archive/wav-files.zip
- Install dependencies with
sudo pip install -r requirements.txt