Forward Backward ResNet.
This work is a continuation of the work done by Cecile Della Valle in https://github.com/ceciledellavalle/FBResNet.
Install requirements:
$ pip install -r requirements.txt
$ pip install .
$ bash ./jobs/test.sh
Datasets
Images
: two sets of images from which are extracted 1D signal datasetsSignals
: dataset constructed from imagesFBRN
: contains FBRestNet files
info.py
: model information. Allows pip installing the FBRN model.main.py
: contains FBResNet class, which is the main class including train and test functionsmodel.py
: includes the definition of the layers in FBRestNetmyfunc.py
: useful functions used such as convolution and Abel operatorsproxop
: contains he cardan class used to compute the proximity operator of the logarithmic barrier
hyperslab.py
: proximity operator for hyperslab constrainthyperscube.py
: proximity operator for cubic constraintscripts
FBRN_click.py
: interface model parameters and options through a click comand line interface. fouriermethod.py
: 1D reconstruction with Fourier filtering method.lowessmethod.py
: 1D reconstruction with Lowess filtering method.optuna_FBRN.py
: explore and find the best hyperparameters for the FBRN model. train_utils.py
: train, test and plot functions.training_time_opt.py
: study FBRN trainig times.notebooks
: collection of jupyter notebooks for different analysis and studies.jobs
: bahs files for submitting model training and evaluation with different parameters.