model.py

Required packages

• Keras
• h5py
• numpy

How to run

• the weights of our trained models exceed 1GB so they could not be uploaded to the repository
• go to main at the bottom of the script
• change the variable data_file to be the path to an hdf5 data file...this file's structure form is specified in the comments of the main and the script make_HDF5.py will create this data file
• the variables label_path and dataset should work as is, but if you want to run the model on the AIC480 dataset or run a multi class model instead, you will have to update them accordingly
• if a model checkpoint named checkpoint.h5 exists in the current directory, it will be loaded -- otherwise a new model will be initialized
• uncomment one of the last two lines in order to train and/or evaluate the model respectively

make_HDF5.py

Required packages

• scipy
• h5py
• numpy

How to run

• this script must be run to recreate the data file as it is approximately 134GB and could not be uploaded to github
• update the variable root to specify the directory of the environment in which the directories data and datasets exist
• will create a data file located at <root>/data/data.h5

regression_labeler.py

Required packages

• numpy

How to run

• this script shouldn't need to be run since the labels are included in this repository
• edit the list variable called paths so that the file path strings inside of it are paths to the training and validation labels for the AIC540 and AIC480 datasets