Salt and Pepper (AKA: series and pictures) - A general structure for deep learning on signals/spectra with PyTorch
This project defines the fundamental structure for work-in-progress deep learning projects aimed at classifying signals by means of the associated time series or the related spectral maps. All the packages required are listed in the requirements.txt file.
To run the basic train procedure, one can use either the script train.py or the notebook train.ipynb. So far the only model implemented is a fully connected network with ReLU activations - the number of hidden layers and units can be changed. The datasets included so far are cinc2017 and ecg_sample and they are automatically downloaded when executing the script/notebook.
To add a new dataset that can be easily loaded, one needs to create a new folder under datasets with the same name as the dataset, add a __init__.py file and implement the following methods:
download_data()- needed only if the data need to be downloaded;load_data()- this method is required and details how to load the actual data, it is supposed to return the series/maps and the associated labels;get_label_names()- this method should return the actual class names associated with the labels.
- comments;
- data class for spectral maps;
- more models;
- more plots;
- ...