Hey. Good to see this repository. We would like to use the same BANet with Keras and Tensorflow. Should we need a separate dataset with boundary information alone? And as per the official implementation, all 3 losses, should be used together. Could you help us understand how to pass two different segmentation masks as labels for training.
Hey. Good to see this repository. We would like to use the same BANet with Keras and Tensorflow. Should we need a separate dataset with boundary information alone? And as per the official implementation, all 3 losses, should be used together. Could you help us understand how to pass two different segmentation masks as labels for training.