P2B
P2B copied to clipboard
Question about losses
Hi, @HaozheQi thanks for your great work! I have a few questions after reading your code carefully.
-
Why did you randomly sample 128 points through FPS when selecting seed, but when calculating loss, you used the first 128 points of the input point cloud as ground truth? I know this question is already appear in #Issue1, but I am still confused in this question.
-
How do you define the ground truth of heading angle (rotation of x-y plane), which cannot be found in your codes clearly but is mentioned in your paper.
Looking forward your reply!
According to the author, the down sample process is random sample, so I think there is no FPS.