TLDR this project aims to classify audios into cough - not cough. To do so, I used VGGish for features extraction and SVM for classification. More details can be found in my reports.
Tested with:
python 3.6
tensorflow 1.14
numpy 1.17
scipy 1.3
scikit-learn 0.21
on Ubuntu 18.04
This project uses a model called VGGish so checkpoints and PCA params are required (for post-processing) when running inference or training SVM:
VGGish model checkpoint
Embedding PCA parameters
Put them in the same directory with README
More details about the model can be found here
NOTED: svm_train.py is out-dated, please use this notebook: SVM_train.ipynb
To run inference (cough recognization) with single files and new records: inference_gui.py (required tkinter)
To run bulk inference from tfrecord file: svm_test.py (remember to change dir_test)