Auditory DeepDream

Clair de Lune by Debussy example

Break beat example

Very Dreamified break beat example

Audio examples using InceptionV3 model

Downloading datasets

Small data set (≈4.6gb):

wget https://www.dropbox.com/s/fxvt2l6bacrya7j/serialized_dataset.zip && unzip serialized_dataset.zip

Running with Docker

Note about GPU usage:

The quickest way to get Tensorflow running on GPU is with a docker container as suggested here. The following are links to the required software and instructions on how to install them.

Requirements for GPU:

• Nvidia Container Tools
• Nvidia Drivers
• Cuda Tool Kit

Build docker images:

chmod +x build_docker_images.sh && ./build_docker_images.sh

Before spinning up the containers, have a look at the docker-compose.yml file. If your machine doesn't have an Nvidia GPU, you will need to comment out the devices being passed into the container. Also set the memory for the container appropriately.

To start up the Jupyter server run:

docker-compose up -d && sudo docker-compose logs -f

Then follow the link with token, example link:

http://127.0.0.1:8888/?token=68fea5fe14cb2a97bb0fb016f0690b91b664e8df9f75436d

To stop the Jupyter server run:

docker-compose kill

Note: Tensorflow's Docker image uses Python3.8