Open jackjansen opened 1 year ago
The point cloud is from Vsense dataset. https://v-sense.scss.tcd.ie/research/vsensevvdb2-v-sense-volumetric-video-quality-database-2/ This also happened for 8i dataset, since we need to scale the point cloud from 8i dataset. (from the python script for loot the scale_factor = 500 )
Discovered while doing the experiments of @zhouzhouha .
For some of the datasets she used it seems that changing
octreeDepth
does not make much of a difference (if any). Even going as low as octree depth 5, when we encoded a 400Kpoint cloud the result was still more than 300K points.@zhouzhouha can you please add a comment here about which dataset it was again? I forgot...
We should investigate whether this is also the case for the 8i datasets. And we should investigate why this happens:
octreeDepth
should be scale-independent. There may be a subtle other issue with the given dataset, maybe it has funny bounding boxes or something?