Hi, thanks for sharing your work.
I tried to reproduce RDFNet results on the NYU Depth V2 dataset using your PyTorch implementation, but i obtained only 23.26% of mIoU.
In a recent paper of yours I noticed that you managed to get a mIoU value of 48.7% with the PyTorch version of RDFNet-101 which is close to the result in the original paper (49.1%).
Could you give me some insight about how you trained the network?
Hi, thanks for sharing your work.
I tried to reproduce RDFNet results on the NYU Depth V2 dataset using your PyTorch implementation, but i obtained only 23.26% of mIoU.
In a recent paper of yours I noticed that you managed to get a mIoU value of 48.7% with the PyTorch version of RDFNet-101 which is close to the result in the original paper (49.1%).
Could you give me some insight about how you trained the network?