IKR
IKR 2018 projekt, audio classification part
External libs:
pip3 install keras
pip3 install tensorflow
pip3 install tqdm
pip3 install librosa
Sources:
- https://blog.manash.me/building-a-dead-simple-word-recognition-engine-using-convnet-in-keras-25e72c19c12b
- https://github.com/manashmndl/DeadSimpleSpeechRecognizer
- https://blog.keras.io/building-powerful-image-classification-models-using-very-little-data.html
Directory structure:
.
├── eval
│ ├── [evaluation data for clasification(could be mixed .png with wav)]
└── my_data
├── nontarget
│ ├── [coppied all nontarget wav files only (nontarget_train/dev dirs)]
└── target
└── [coppied all target wav files only (target_train/dev dirs)]
by-product od execution audio_classification.py will be two numpy arrays
- nontarget.npy
- target.npy