search

mayhemsloth / Drum-Tabber

Automatic drum tab project using self-created data set and TensorFlow NN
6 stars 1 forks source link

Implement new data repository structure and Class for importing data #10

Closed mayhemsloth closed 4 years ago

mayhemsloth commented 4 years ago

Following Matt's input on the recommended data repository structure, I want to restructure how I store my data so that it will be easier to grab, process, and store things in memory when I have a larger data set. Additionally, I should make my own Class that is used to import the data properly, so I can reuse that variable.

He recommended I do the following. Put all the songs into their own folder in a Songs folder, with the tabs and metadata being the same name in every folder: Songs/ |--- example_song_1 |--- example_song_1.mp3 |--- metadata.json |--- tabs.txt |--- example_song_2 |--- example_song_2.mp3 |--- metadata.json |--- tabs.txt

I like this structure better, but I don't like naming the metadata and tabs as these exact copy of names because then it can become confusing for me personally when organizing the data for what file belongs to what. If the song file can be the same name as the folder and the code runs just fine, then surely the .json and .txt files can also be named based on the file name so

mayhemsloth commented 4 years ago

I ended up implementing the new data repository but didn't make a Class for the importing. I just used a normal Python function that batch processed everything based solely on the subdirectory file path of the Songs folder.