As the paper described, the dense metric depth map D_m could be got from D_m = f_x / w / c.
When training the model with datasets, like vkitti2 or ISR, how could I get the canonical inverse depth (ground truth) from the given depth in the datasets?
Regards,
As the paper described, the dense metric depth map D_m could be got from D_m = f_x / w / c.
When training the model with datasets, like vkitti2 or ISR, how could I get the canonical inverse depth (ground truth) from the given depth in the datasets?
Regards,