The goal is to recover the 3D coordinates of pixels in GSV images.
We will use MegaDepth to get the depthmap.
We want to check the following :
pinhole camera model
Megadepth model is accurate enough for our purpose -> "Flooding up to level x" masks seem correct
Try the approach using ground truth depth maps
(Bonus : can we find the scaling factor to go from relative distances to metric distances between points ? )
The goal is to recover the 3D coordinates of pixels in GSV images. We will use MegaDepth to get the depthmap. We want to check the following :