How brotherz11/tfg_24 works
Project type and purpose
• This is a full-stack, Python-based web application (a “TFG” or capstone project) built on Flask.
• Its primary function appears to be audio transcription and evaluation. The app offers three transcription models (basic, onsets, “transkun”) and uses Magenta’s open-source music/ML library for underlying model implementations.
File structure
• Top-level:
• app.py – Flask entry point
• app/ – your application package
• environment.yml, requirements.txt – dependency definitions
• magenta/ – vendored Magenta library (hundreds of ML models, pipelines, utilities)
• app/ contains:
• models/ – wrappers for three transcription approaches
• utils/ – preprocessing (spectrograms, audio cleaning), downloading, evaluation, scoring
• routes.py – HTTP endpoints
• static/ & templates/ – CSS and the index.html UI
• No database folder is present – likely no persistent datastore.
README
• Minimal: “Este es mi proyecto de TFG…” – no additional architecture notes.
• We infer architecture solely from file layout and naming.