An automatic beatmap generator using Tensorflow / Deep Learning.
Demo map 1 (low BPM): https://osu.ppy.sh/beatmapsets/1290030
Demo map 2 (high BPM): https://osu.ppy.sh/beatmapsets/1290026
https://colab.research.google.com/github/kotritrona/osumapper/blob/master/v7.0/Colab.ipynb
For mania mode: mania_Colab.ipynb
https://github.com/kotritrona/osumapper/wiki/Complete-guide:-creating-beatmap-using-osumapper
- Refer to https://github.com/kotritrona/osumapper/tree/master/v6.2 for version 6.2
- Refer to https://github.com/kotritrona/osumapper/tree/master/v7.0 for version 7.0
Don't train with every single map in your osu!. That's not how machine learning works!
I would suggest you select only maps you think are well made, for instance a mapset that contains all 5.0 ~ 6.5☆ maps mapped by (insert mapper name).
- I have made a maplist generator under
v7.0/
folder. Runnode gen_maplist.js
under the directory to start. - the other way to create a maplist.txt file to train the model is by using the maplist creator.py script (found in v6.2 folder). running this should overwrite the maplist.txt in the folder with a new one using the maps from the collection folder you have specified.
- Rhythm model
- CNN/LSTM + dense layers
- input music FFTs (7 time_windows x 32 fft_size x 2 (magnitude, phase))
- additional input timing (is_1/1, is_1/4, is_1/2, is_the_other_1/4, BPM, tick_length, slider_length)
- output (is_note, is_circle, is_slider, is_spinner, is_sliding, is_spinning) for 1/-1 classification
- Momentum model
- Same structure as above
- output (momentum, angular_momentum) as regression
- momentum is distance over time. It should be proportional to circle size which I may implement later.
- angular_momentum is angle over time. currently unused.
- it's only used in v6.2
- Slider model
- was designed to classify slider lengths and shapes
- currently unused
- Flow model
- uses GAN to generate the flow.
- takes 10 notes as a group and train them each time
- Generator: some dense layers, input (randomness x 50), output (cos_list x 20, sin_list x 20)
- this output is then fed into a map generator to build a map corresponding to the angular values
- map constructor output: (x_start, y_start, vector_out_x, vector_out_y, x_end, y_end) x 10
- Discriminator: simpleRNN, some dense layers, input ↑, output (1,) ranging from 0 to 1
- every big epoch(?), trains generator for 7 epochs and then discriminator 3 epochs
- trains 6 ~ 25 big epochs each group. mostly 6 epochs unless the generated map is out of the mapping region (0:512, 0:384).
- Beatmap Converter
- uses node.js to convert map data between JSON and .osu formats
If you want to cite osumapper in a scholarly work, please cite the github page. I'm not going to write a paper for it.