Depth Completion Experiment Details

[zuoym15]

Hi,

I have a question regarding the Tab2. in the paper. Could you help me clarify the experiment settings:
The baselines you compared against (e.g., CompletionFormer) are tested with 500 points on the 228x304 resolution. This corresponds to 0.7% density. However, in your paper you mention "During training and evaluation, only 2 % of ground-truth pixels are available.". Do you use denser points during testing on NYU compared to baselines or I misunderstood?
Also thanks for releasing the code! Would it be possible for you to release the depth completion part as well? I'm curious to try how it works on my custom datasets.

Best