Open ccj5351 opened 2 years ago
Hi, thanks for your interest in the project!
That is a good point, there does seem to be a mismatch between the paper and the code. In reality it shouldn't matter too much - both depth networks will be in the same arbitrary scale as the pose network due to the reprojection loss, and thus their min/max depths will be very similar (although the student network with access to the cost volume should be more accurate).
If you planned to run an experiment, I'd be interested in hearing if changing to using the student network affects the scores at all.
Thanks again!
Sure. Thanks for your reply. Will let you know when some results are available.
Thanks for sharing your amazing work and code.
Have a question about the update_adaptive_depth_bins() function (Line364 around). It is mentioned in your paper, the depth range is dynamically updated by min and max of MVS depth (i.e., the student network). When checking the code, the mono_depth is used instead. Do I misunderstand that? Or the MVS depth will be learned to mimic the Mono depth? Thanks for your clarification.