Img2Music is a network that generates genre-specific music snippets from the features of an input album cover image.
This submission contains 3 files:
Model_Showcase.ipynb- Demonstrates the model in action, loads pre-trained weights, no training required.Full_Training_Pipeline.ipynb- Code for obtaining weights for each part of the model. Estimated training time using A100 gpu: ~2-3 hrs.report.pdf- Concisely describes project goals, methodologies, results & interpretations, and contains all references.
I recommend starting with Model_Showcase.ipynb or the report to understand the model itself and the context of this project.
Full_Training_Pipeline.ipynb showcases all of the code and approaches mentioned in the report.
Note that the .ipynb notebooks designed to run in Google Colab.