search

facebookresearch / TemporallyConsistentDepth

Code for our CVPR 2023 paper on online, temporally consistent depth estimation.
Other
86 stars 3 forks source link

Details for reproducing the tables. #6

Closed kami93 closed 10 months ago

kami93 commented 1 year ago

Hi. Thanks for your efforts.

I am trying to reproduce the main table 1 reported in the paper, however, I am struggling to find which exact sample names and frame durations are used for your evaluation. The paper reports that:

We evaluate the monocular variant of our method on 20 60-frame sequences from different scenes.

Perhaps more details including the sample names and the frame duration could be shared? Thanks!

nkhan2 commented 11 months ago

Hi kami93, Thank you for your comment. We use the first 60 frames (0.jpg to 59.jpg) from the first 20 scenes of the ScanNet dataset (scene0000_00 to scene0019_00) for our evaluation.

nkhan2 commented 10 months ago

I believe this issue has been resolved.