Pretrained semantic segmentation model's results look strange

zju3dv / manhattan_sdf

Code for "Neural 3D Scene Reconstruction with the Manhattan-world Assumption" CVPR 2022 Oral

https://zju3dv.github.io/manhattan_sdf/

Other

491 stars 35 forks source link

Pretrained semantic segmentation model's results look strange #16

Closed kecol closed 2 years ago

kecol commented 2 years ago

I've downloaded the pretrained model.pth file and used cuda 10.1, pytorch 1.6 and detectron 0.4 as it was required but the results on simple images look not good. Do you know what might be wrong here?

ghy0324 commented 2 years ago

Hi! We also noticed that the semantic segmentation predictions are not good, you can refer to the semantic maps in our processed data. And that is one of the reasons why we design the joint optimization strategy.