You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
After getting no help on the Mask-RCNN, I wanted to give a try to FCN semantic segmentation again with my coco-style dataset. Trained the network for 20 epochs, 64 image size and with 3 classes (background + 2 classes), however, received the following result, which makes absolutely no sense whatsoever.
I am not sure what the problem is in both implementations in the repository, I guess they were tested only with shapes dataset, and they don't work with others. Or perhaps the loading function does not generate the masks properly.
Any idea?
The text was updated successfully, but these errors were encountered:
After getting no help on the Mask-RCNN, I wanted to give a try to FCN semantic segmentation again with my coco-style dataset. Trained the network for 20 epochs, 64 image size and with 3 classes (background + 2 classes), however, received the following result, which makes absolutely no sense whatsoever.
I am not sure what the problem is in both implementations in the repository, I guess they were tested only with shapes dataset, and they don't work with others. Or perhaps the loading function does not generate the masks properly.
Any idea?
The text was updated successfully, but these errors were encountered: