I am trying to reproduce your reported evaluation scores on NYUv2 and SUNRGBD dataset. I used MiT_B5 model but the results I got are far less than what you claimed in the paper. In NYUv2, I got only 51.9 meanIOU compared to 57.7 in the paper. In SUNRGBD, I got only 49.5 compared to 53.3 in the paper. Furthermore, the evaluation score printed by your code is confusing to me. Why do you have 3 different IoU scores for (RGB), (depth), (ens) separately? What is (ens) by the way?
Hi,
I am trying to reproduce your reported evaluation scores on NYUv2 and SUNRGBD dataset. I used MiT_B5 model but the results I got are far less than what you claimed in the paper. In NYUv2, I got only 51.9 meanIOU compared to 57.7 in the paper. In SUNRGBD, I got only 49.5 compared to 53.3 in the paper. Furthermore, the evaluation score printed by your code is confusing to me. Why do you have 3 different IoU scores for (RGB), (depth), (ens) separately? What is (ens) by the way?