I have 3 questions:
-- To convert depth output to PNG, you multiply by 256. Why?
-- What units are the metric depth output? meters?
-- What method do you recommend to scale up the depth results to an input image size?
I don't know the reason but you could refer to here. I just follow the implementation.
meters, yes.
As far as I see, if you adopt a reasonably good tiling strategy, the depth result would have the same resolution as the input. It might be not so necessary for an extra upsample. That's the meaning of PatchFusion
I have 3 questions: -- To convert depth output to PNG, you multiply by 256. Why? -- What units are the metric depth output? meters? -- What method do you recommend to scale up the depth results to an input image size?
Thanks