Strength of supervision required for training

[IssamLaradji]

Does this method require the full semantic segmentation per-pixel labels or the instance segmentation per-pixel labels ?

Thanks!

[jponttuset]

Only instance segmentations. I would, however, advise you to check our newest version: Convolutional Oriented Boudnries (http://www.vision.ee.ethz.ch/~cvlsegmentation/cob/code.html) Best,