resnet_cough_classifier

System Requirement

Install Conda
Need Python versions >= 3.7.6
Install TensorFlow: pip3 install tensorflow
Install Keras: pip3 install keras
Install scikit-learn: pip3 install scikit-learn
Install pandas: pip3 install pandas

Steps to run the inference

Download the ResNet model into the resnet directory and run python3 prediction_generator.py --model resnet. The script will run this network through the test data and produce two outputs: a classification reports and a confusion matrix, both stored in resnet

Step to run the end-to-end pipeline

Download the raw audio signals into raw-data directory (since the sizes of the raw data is quite large, so they are not included in this repository). For access to the raw data, please contact the FluSense authors.
To create processed spectrogram images from these raw audio signals, run python3 processed_data_generator.py
To create the train-val-test datasets, run python3 label_generator.py
(Optional) to look at the class distribution , run python3 stat_generator.py
To train the neural network classifier, run python3 model_generator.py --model resnet. IMPORTANT NOTE: the code also provide the ability to train other well-known deep learning architectures such as DenseNet, Xception and InceptionV3, and can be obtain by passing densenet, xception, inception to the argument --model.
To test the classifier, run python3 prediction_generator.py --model resnet. Two CSVs are output: a classification reports and a confusion matrix in the directory named after the model used.