Content-based filtering for music recommendation system

Description:

This server is using CNN, RNN and CRNN models to classify music by genre & to create personal recommendations utilising either cosine distance or Jensen-Shannon distance. Part of Anitra music recommendation system.

Installation (for dev purposes):

1. Install dependencies

pip(3) install -r requirements.txt

2. Install ffmpeg (don't forget to add link to it to your PATH env variable)

3. Run

python index.py

known issues

if you are getting Tensor Tensor("dense_2/Softmax:0", shape=(?, 10), dtype=float32) is not an element of this graph. error, try adding model._make_predict_function() after loading the model