DORN_pytorch

DORN_pytorch copied to clipboard

Regression layer mismatch on train and test times

2

It seems that while training, the regression prediction tensor is reshaped into [B, C, H, W]. Whereas in test time, only half of the output tensor is kept, resulting in...

evinpinar

question about kitti depth

1

on the kitti depth dataset site,should we also download the projected raw lidar scans,or will the annotated depth map dataset link be enough for evaluating this project?

sparshgarg23

the value of loss

1

How much is normal for dorn loss? I found it much larger than any other regular loss. And it may depends on alpha or beta?

EE102-JN

Hi, have you been able to replicate the numbers reported in the paper with this codebase? If so, can you please share a pre-trained model? Thank you!

VitorGuizilini-TRI

evinpinar

How to run this code, I did not find the entrance of this code, please help me

1

thanks for your great work

ShihuaiXu

Hi, thank you for your contribution. Following the BerHuber loss that u implemented, I am not sure if that is the correct implementation. I would suggest something like this: class...

jrodriguezpuigvert

line 75 labels = labels.float()

LJNDDUP

There was an incorrect flag in original monodepth lidar projection which was kept there for reproducibility purposes. Using incorrect flag will calculate depth from lidar not the camera. see github.com/mrharicot/monodepth/issues/166#issuecomment-399392115

jahaniam

Hi, I am unsure how to run this code, as no instructions are provided. What should I do after cloning the repo for running the code? Or do I have...

AJAY31797