converting predicted disparity to depth value

[Torment123]

Hi, I'm applying the KITTI pretrained FastACV on a new dataset (KITTI360), and trying to convert the predicted disparity of stereo pair to depth value, and the process I do is as below:

given a stereo pair of KITTI360 dataset, first compute the predicted disparity D using fast ACV.

Then compute the corresponding depth map using the formula Depth = f x B / Disparity (1), where f (unit:meter) and B are focal length and baseline of the rectified camera pairs setting in KITTI360 dataset. I'm not sure whether I'm missing any intermediate steps, because although the predicted visualization looks good, but the depth metric distance computed based on (1) and ground truth depth is very large. I'm appreciate it if you could help me pointing out where I did wrong, thanks

[Nimisha-Pabbichetty]

were you able to figure this out?