Skip to content

Latest commit

 

History

History
32 lines (23 loc) · 1.16 KB

README.md

File metadata and controls

32 lines (23 loc) · 1.16 KB
Raw

tfgenre - Vlady Veselinov

Download and unzip trained models in root folder (2GB)

To evaluate accuracy of all models:

sh evaluate.sh

To train all models

sh train.sh

Optional audio data extraction guide with Essentia (macOS only)

Data set used for this project can be found here

  1. Install Essentia with Homebrew: brew tap MTG/essentia && brew install essentia

  2. Make a data/audio folder in the project root containing subfolders of audio files. Subfolder names used as categories, i.e. data/audio/car_horn/11251.wav

  3. Run src/extract_features.py with Python2.7 to load audio from data/audio and produce JSON files with audio features

  4. Run write_csv.py to distill all JSON to one .csv

Optional Run TensorBoard for visualisations:

Download and unzip log files in root folder

Start TensorBoard

sh tensorboard.sh