rgb_loss stuck at zero

[iernstig]

Hi! Thank you for the great work, I'm trying to get system to work on my own images, but am facing some difficulties. I obtain the position information using an aruco board on which I place my object. Now, the algorithm moves past the ray-tracing step properly, but then the algorithm outputs an rgb-loss of "0.0". Did anyone encounter something similar? I am very grateful for any type of response :)

[za-cheng]

Hi,

Were you able to solve this problem? I came across the same issue with rgb_loss always evaluates to 0 on other images. More specifically, the network_object_mask in rgb loss function seems always zero everywhere.

In preparing the images I arranged them in DTU style with images and masks. And populated cameras.npz with world-to-image projection matrix and object-to-world scale_mat, with the object coordinates confined in a unit sphere.

Edit: I just realised I had used the wrong coordinate system for camera projection. After correcting this the rgb_loss becomes non-zero.

[qinzhibao123]

@za-cheng
Hi! How did you finally solve this problem? What is the correct coordinate system for camera projection? How do I get the initial world_mat in cameras.npz?

[JasonW-00]

Hi,

Were you able to solve this problem? I came across the same issue with rgb_loss always evaluates to 0 on other images. More specifically, the network_object_mask in rgb loss function seems always zero everywhere.

In preparing the images I arranged them in DTU style with images and masks. And populated cameras.npz with world-to-image projection matrix and object-to-world scale_mat, with the object coordinates confined in a unit sphere.

Edit: I just realised I had used the wrong coordinate system for camera projection. After correcting this the rgb_loss becomes non-zero.

Hi！Could you tell me in detail which part of your negligence caused the calculation to be incorrect? I also faced the same problem and it bothered me for a long time