---

Training of IA models to generate music in MIDI format, resembling old NES videogames sountracks.

Project for Deep Learning course a.y. 2023-2024.

Team:

👨‍💻 Valerio Di Zio - valerio.dizio@studio.unibo.it

👨‍💻 Francesco Magnani - francesco.magnani14@studio.unibo.it

👨‍💻 Luca Rubboli - luca.rubboli2@studio.unibo.it

Available code

In this repository we present two possible approaches for generating MIDI music, each with its own Jupyter Notebook.

Transformer approach:

GAN approach:

Useful links

Here's some useful links:

Artifact	Link	Size
NESGEN_Maestro15.zip	https://drive.google.com/file/d/1qoTIkmhmj0z4854FFofz02KLk5DnyqB1/view?usp=drive_link	175 MB
NESGEN_Nesmdb5.zip	https://drive.google.com/file/d/1FfofRGCzuD6x54Q3mF2c_f50wuzbPp_i/view?usp=drive_link	175 MB
NESGEN_gan50	https://drive.google.com/file/d/1XN71Mnc3C4gzr1w8u5xgcOAsX9zNds-m/view?usp=drive_link	1.54 GB
maestro_tokenizer.json	https://drive.google.com/file/d/1SDRkoWwyuSl4udoCHdcitjLLm9d0kfxS/view?usp=drive_link	1.6 MB
nes_tokenizer.json	https://drive.google.com/file/d/1ul-khIle1ue9oUGqb1Z7qHxk7LSuxrWw/view?usp=drive_link	1.9 MB
gan_tokenizer	https://drive.google.com/file/d/1Uf734gntq6RLpAvcruWuKcHqPmcZ_4O8/view?usp=drive_link	212 KB
all_ids_maestro_train.txt	https://drive.google.com/file/d/1IQToXD9s8g4L-AlK-MY4qvGoLZ-p7bMw/view?usp=drive_link	1.19 GB
all_ids_maestro_valid.txt	https://drive.google.com/file/d/1DWjViUKpW07LfbGimlhhhGdK7oQaJpj-/view?usp=drive_link	545 MB
all_ids_maestro_test.txt	https://drive.google.com/file/d/1ow5MhMc_Lm3EdZSmH9BYUJfDELtMQGQl/view?usp=drive_link	272 MB
all_ids_nes_train.txt	https://drive.google.com/file/d/1iVwuKKT2sIezyVIDB9STL--FYZf0XcxF/view?usp=drive_link	122 MB
all_ids_nes_valid.txt	https://drive.google.com/file/d/1FoOC3x4GbdDe0fwlzE1Zs8fqZXyFQCIA/view?usp=drive_link	57 MB
all_ids_gan_train.txt	https://drive.google.com/file/d/1VU4ngttqwACFEtnv7l6B8xG2Z1eS3GWZ/view?usp=drive_link	1.6 GB
all_ids_gan_valid.txt	https://drive.google.com/file/d/1c8hgL4J2BVS4_XXuZNlkuTRYEHduZTdy/view?usp=drive_link	104 MB
NESGEN_maestro_100generations	https://drive.google.com/file/d/1ToDyVJRTVaxmYSfiGPmC-KnkI57M6i76/view?usp=drive_link	57 KB

To download these files inside a Colab notebook you can simply use the gdown command like this:

# For example, the NESGEN_Maestro15.zip file. Simply replace the ID with the part of Google Drive URL after 'd/' until '/view'
!gdown 1qoTIkmhmj0z4854FFofz02KLk5DnyqB1