Mnist Tools

mnist_seg.py

A script for generating the mnist segmentation dataset, mnist segmentations are useful when one requires quick segmentation model convergence, i.e. for hyperparameter testing.

Run with

python mnist_seg.py --t mask_threshold (int) --s save_path (str) --p pkl_path (str) and optionally, --RGB (Bool)

mnistm_gen.py

A script for generating the mnistm dataset, which uses the mnist segmentations generated with mnist_seg.py as ground truth.

Run with

python mnistm_gen.py --mp mnist_path (str) --ip imgnet_path (str) --s save_path (str)

mnistmPytorchLoader.py

A script to load mnistm into pytorch using the existing dataloader class. You should use this script as an auxilary to your pytorch model and train.py.

Run with

python mnistmPytorchLoader.py