Camera model used for metric depth training

[RuibinMa]

Dear author,
Did you use images of any particular camera model (i.e.. perspective, xxx focal length, etc) to train the metric depth models? I noticed when applying your metric depth model on different images (outdoor phone images for example), the depth map accuracy can vary a lot. I guess that might be related to the different camera models between my testing and your training so the model couldn't figure out the missing scale factor.
Would it be possible for you to share some details of the camera models of your training images for the metric depth models?
Thank you very much!

[crysoberil]

I would also love to get get clarity on this from the authors!

[LiheYoung]

We use Hypersim and Virtual KITTI 2 as our training data for indoor and outdoor metric depth models, respectively. You may refer to these two datasets for details of their camera configurations.