A very simple python app to detect the genre of an audio file using a Mel Spectogram and a simple CNN with PyTorch.
Currently we're trained with just 10 Geners using the GTZAN data set, but working to retrain on the Million Songs dataset for a wider array of results and geners. As shown here http://millionsongdataset.com/sites/default/files/AdditionalFiles/unique_terms.txt
$ docker-compose up -d application
Then just send it a file.
curl --request POST \
--url http://127.0.0.1:5000/ \
--header 'content-type: multipart/form-data; boundary=---011000010111000001101001' \
--form file=/path/to/file
Million Songs GTZAN Genre Collection
Big thanks for cetinsame for the work on the original model.