lang_classify

A nn model detacting an audio is Taiwanese or Chinese.

Requirement

• ffmpeg
• youtube-dl

Setup

Preprocess

Requirement: Mno .wavfile with sample_rate = 8192.

ps: preprocess/collect.py, preprocess/convert.py and preprocess/convert.sh may be helpful.

Using audio in wav/ building dataset in feature/.

$ ./preprocess/build_feature 

Training

Set your data size in train.py.

Trining model:

$ ./train.py

Tensorboard:

tensorboard --logdir=logs/

Evaluation

$ ./eval.py