Closed austinmw closed 6 years ago
So somebody found that this isn't a bug it the data apparently it's a bug in skimage.io.imread
called from mrcnn/utils.py
which can't correctly output large images' shape: https://github.com/matterport/Mask_RCNN/issues/525
I'm getting this error very intermittently during training (approx. 10 times per 1000 iterations). I have variable size images and masks, so I'm thinking this may be an issue with some of the very large images in my dataset (for example, sizes 5456x3632, 2592x1944, etc.). It continues to train without crashing due to the error, but I'm unsure if there will be any negative consequences later on.