mrharicot / monodepth

Unsupervised single image depth prediction with CNNs
Other
2.21k stars 628 forks source link

world coordinates #256

Open KunikaValecha opened 3 years ago

KunikaValecha commented 3 years ago

I suppose, the depth produced from disparity map is in world coordinates. Then given calibration matrix, how can I generate coordinates in world frame? My approach is to make homogeneous vector from (xpixel, ypixel) and then multiply it with inverse of calibration matrix. That will That will give us ratio of x/z, y/z. Further ratio * z(from disparity) will give xworld, yworld. Is this correct?