amoszhu / cvpr2019 Goto Github PK

Hello
very interesting work
In the horse_disparity_... matrices, the intrinsic matrices are missing in the camera parameters
can you provide them ?
Thanks in advance

Hello, Dr. Zhu.

Thank you very much for your excellent work in open source. I recently used kinect v2 and FLIR polarizing cameras to get data and test your dem. It turns out that the result is not as ideal as the paper, and I observe that the polarization image data provided by you is more accurate. Therefore, I would like to ask you the following experimental questions:

1. What kind of experimental equipment or operation can ensure the accuracy of polarization information collection?
2. What material or spray paint do you use as the surface reflective material of the object.
3. Which light source device do you use as a point light source? At the same time, does the environment need absolute light isolation?
4. Can you provide more of your test data?

If you are free, please help me.

Greetings.
Wang Ya.

Hello, I followed the process in the paper and first obtained a rough depth map with a binocular camera, but I have doubts about the step from rough depth map to guiding surface normal. The python code in this part of normal correction is to directly take the guide surface normal as input, and I am not very clear how to flush the rough depth map to solve the normal, I have two ideas (1) according to the orthogonal projection model, use a simple forward difference to find the gradient and then find the normal, (2) or according to the transmission projection model to find the gradient and normal? Looking forward to your answer, thank you