Thank you for sharing this wonderful work. I really appreciate it.
In the prediction, there are fcn_outputs (value from 1 to 132) and panoptic_outputs (value from 1 to 255), could you tell me what is the difference between them? In most of cases, they are almost the same, only a few pixels difference.
And what is the meaning of mask_probs with shape (81,28,28)? It represents instance category? How to map it back to original image?
Dear UPSNet Group:
Thank you for sharing this wonderful work. I really appreciate it.
In the prediction, there are fcn_outputs (value from 1 to 132) and panoptic_outputs (value from 1 to 255), could you tell me what is the difference between them? In most of cases, they are almost the same, only a few pixels difference.
And what is the meaning of mask_probs with shape (81,28,28)? It represents instance category? How to map it back to original image?
I am using CoCo resnet-101 and test on some indoor images from web. Code is here https://github.com/betterze/tmp/blob/master/UPSNet%20Prediction .
Thank you very much for your help.
Best Wishes,
Alex