Best Beat Analyzer

Problems

Beatmaps are very tedious to put together. To ensure quality, human input for detailed beatmap generation is required (which can take many iteration and it is a slow/repetitive process) Traditional algorithm generated beatmap can be skewed by factors such as volume/amplitude peaks, noise (in the wave forms). We want to implement a better, more accurate and automatic beatmap generator

What it does

Visualize based on music file and side-by-side beats to match with the music pattern comparson.

How we built it

Using Unity3D engine for visualization and front end. Osu! beatmap data for truth/training data. Google Tensorflow machine learning library for training and learning.

Challenges we ran into

First time using Tensorflow library for machine learhning, extraction of data original song forms and beats from Osu!'s beatmaps.

• hard to gather and set training threshhold and criteria.

Accomplishments that we're proud of

-Able to utilize tensorflow to start running training algorithm -create and visualize music data

What we learned

-Tensorflow -Data extraction -Data Analysis -Data visualization

What's next for Best Beat Analyzer

-Hoping to create better algorithm in future, better training material, better music analysis

Built With

• c#
• osu
• python
• tensorflow
• unity