search

Morpheus3000 / PIE-Net

Official model and network release for my CVPR2022 paper.
38 stars 8 forks source link

CCR image visualization #7

Open yuhaoliu7456 opened 1 year ago

yuhaoliu7456 commented 1 year ago

Can you share the code to obtain fig-(b)? I tried your code of ratioCals Class, but it lacks a function in the forward function.

Morpheus3000 commented 1 year ago

Hi, thanks for your interest in the work. Sorry for the later reply, I was caught up.

As for the code, have you checked the network.py file? There is forward functions for the code, since that is used to calculate the ratios. Specifically, L84. And you can see an example of how the network calls it from L211.

Good luck!

yuhaoliu7456 commented 1 year ago

Hi, thanks for your interest in the work. Sorry for the later reply, I was caught up.

As for the code, have you checked the network.py file? There is forward functions for the code, since that is used to calculate the ratios. Specifically, L84. And you can see an example of how the network calls it from L211.

Good luck!

Hi thanks for your reply, I tried your method, but still can not get the same figure. For example, the CCR_img of the fig "uvc_camera_cam_0_f00140_undist.png" is

image
Morpheus3000 commented 1 year ago

Hi, sorry for the late response, I was caught up. The image looks reasonable. The output like this is expected, since the image is from the realworld on a very low quality camera. So the image captured has a lot of noise and sensor artifacts. The image in the paper are different, because they are synthetic and thus lack the instability and uncertainties in the pixels that you see here. That is the reason why this information by itself doesn't give the total information and we need the proposed algorithm to have a recover and compensate for that.