Closed tizianoGuadagnino closed 4 months ago
The difference in performances between the two versions is nothing, and I feel the code for the random sampling + robin_set
is much more at the end, and it is not worth the space. We will keep the (modernized) original version (hopefully #362 ) and think about this later.
This PR tries to maintain the classical behavior of our beloved
VoxelDownsample
while taking what is good from #347, a.k.a. avoiding the doublefor
loop over the LiDAR frame and hash table.All the credit for this change goes to @l00p3. Once it is ready to merge (if), I will ensure you are between the contributors.
I will start the testing process today and come back to you, @nachovizzo @benemer feel free to be
-Wpedantic