How to get the metric depth?

[kay-Lifeng]

Dear author:
Thank you very much for your contributions in this paper！
I try to get the depth in evaluation. Theoretically, we can get the metric depth between min_depth and max_depth(meter) after the function 'disp_to_depth' . However, I observed several groups of depth generated in this way, it seems not like a correct metric depth. As below:
(Pdb) pred_depth
array([[0.7076706 , 0.7073359 , 0.70599675, ..., 0.60197115, 0.59946537, 0.5988389 ],
[0.70776784, 0.7074405 , 0.7061312 , ..., 0.60220945, 0.599669 , 0.59903383],
[0.7082062 , 0.70791256, 0.706738 , ..., 0.603285 , 0.60058796, 0.5999137 ],
...,
[0.11618532, 0.1162222 , 0.11636975, ..., 0.11287601, 0.11278087, 0.11275709],
[0.11617843, 0.11621682, 0.11637037, ..., 0.11286871, 0.11277094, 0.11274651],
[0.1161769 , 0.11621563, 0.11637051, ..., 0.11286709, 0.11276875, 0.11274416]], dtype=float32)
The depth number is very small compare to real depth, much of them less than 1.0. So how to get the metric depth?

[weiyithu]

Hi, which model do you use? For the scale-ambiguous model, it cannot predict real-world scale. However, for the scale-aware model, it is able to predict real depth.

[rockywind]

Hi，why scale-aware model's Abs Real is larger than acale-ambiguous in nuScenes dataset?

[weiyithu]

For scale-ambiguous experiments, we conduct median-scaling during evaluation. If scale-aware depths are absolutely accurate, the scale-aware results should be equal to scale-ambiguious results. However, currently scale-aware depths have errors and median-scaling will boost the quantitative results.

[rockywind]

Thank for your help!

[v-fedoseev]

Hi, which model do you use? For the scale-ambiguous model, it cannot predict real-world scale. However, for the scale-aware model, it is able to predict real depth.

Does the scale-aware model pre-trained on nuScenes actually produce the real depth? Could you please describe how do I get the real depth out of the produced images? It would be very helpful, thank you!