Why does the depth value of the depth map generated by the rendering not match the actual one

[qinshihao12]

Thank you for your excellent research, but I have a question for you here.
I tried to train and render a real scene that I captured myself, but I found that the depth values of the distance_mean_.tiff and distance_median_.tiff plots obtained if rendered are about three or four times different from the depth values of the real scene captured when read using cv2.imread, so can you tell me the so can you tell me the reason for this?
And what values should I enter when I want to use compute_disp_metrics?