Paul Alan

Try to modify num_workers to 0 in train_semseg.py. You can find details in https://github.com/matterport/Mask_RCNN/issues/93 It works fine in my PC.

tf=1.13.1 & keras=2.1.4