Negative scale in depth matching

[turandai]

Hi, thanks for this fantastic work, I am adopting your code to solve similar problem. And there is an issue that, the compute_scale_and_shift() function returns a negative scale value which flipped the relative depth information between pixels. I assume this is not expected to happen. I wonder have you met similar problem in your expirements? or maybe some clues about the potential cause of this problem. Thank you very much!

[HLinChen]

I meet the same problem.

[niujinshuchong]

Hi, can you check if your depth map output is disparity?