Code made for JVTP in association with Nvias.org
- mainly made according to describrion in https://arxiv.org/pdf/1505.04597.pdf
- something inspired in https://keras.io/examples/vision/oxford_pets_image_segmentation/
- something more inspired in https://github.com/nikhilroxtomar/UNet-Segmentation-in-Keras-TensorFlow/blob/master/unet-segmentation.ipynb
but everything made to work with simple custom dataset pictures need to be .jpg masks need to be .png with mask values 1,2,3...important!!! check before for more than one mask need to repair some functions in train.py and DatasetGenerator.py made by Filip Jašek 26.7.2021 github: Fiiila