This is a repository for everything related to my bachelor's thesis. It continues the work started in my seminar project.
Most scripts require tensorflow and numpy.
The train.py script begins the training of a new model with the architecture proposed in this thesis. It accepts four arguments. The first one is the path to the data folder generated by the merge_datasets.py script. The other three arguments are either True or False and they toggle different settings in this order - greyscale, histogram_normalisation, data_augmentation.
Example usage:
python3 train.py datasets/data/ True False False
# for a greyscale model with no histogram normalisation and no data augmentation
The label_image.py script loads a model and tries to classify the provided image.
Example usage:
python3 label_image.py
# to label the provided example_sign.jpg
python3 label_image.py --image=another_image.jpg
# to label a different image
python3 label_image.py --model=path/to/another_model.h5
# to load a different keras model for the classification
python3 label_image.py --grayscale=True --normalize=True
# converts the labeled image to grayscale and applies histogram equalization
python3 label_image.py --help
# for more information and options