search

isl-org / MiDaS

Code for robust monocular depth estimation described in "Ranftl et. al., Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer, TPAMI 2022"
MIT License
4.25k stars 597 forks source link

Output Depth maps #227

Open Sangipc opened 11 months ago

Sangipc commented 11 months ago

I have a few doubts, if you could kindly clear them

  1. Could you clarify the number of layers in the depth map produced by the MiDaS model and how these different layers relate to the depth predictions?

  2. What is the unit of the depth maps which are produced as output ? (pixel, mm, cm,...)?

  3. How can we transform the output of the MiDaS model, which contains relative depth values, back to the original input resolution? This would help establish a pixel-to-depth map relation between the input image and the corresponding depth map.

elenacliu commented 11 months ago

I have the same question

Halloweenpink commented 7 months ago

output depth map is relative, I tried with my camera as input, same object depth changed as I rotated around it( cuz the changing of background)