In Segment Anything in High Quality, do model recognize and learn the gt mask by class?
If you look at the code:
class OnlineDataset(Dataset)
in train/utils/dataloder.py:
if len(gt.shape) > 2:
gt = gt[:, :, 0]
if len(im.shape) < 3:
im = im[:, :, np.newaxis]
if im.shape[2] == 1:
im = np.repeat(im, 3, axis=2)
Looking at this code, it looks like the gt mask is changed to a black and white image even though it is divided into colors by class. Does it have no effect on accuracy even if the model does not learn class information?
In Segment Anything in High Quality, do model recognize and learn the gt mask by class?
If you look at the code: class OnlineDataset(Dataset) in train/utils/dataloder.py: if len(gt.shape) > 2: gt = gt[:, :, 0] if len(im.shape) < 3: im = im[:, :, np.newaxis] if im.shape[2] == 1: im = np.repeat(im, 3, axis=2) Looking at this code, it looks like the gt mask is changed to a black and white image even though it is divided into colors by class. Does it have no effect on accuracy even if the model does not learn class information?