input_pointcloud_n #5
Description
Although the original dataset contains one hundred thousand points per frame, is there a reason only 300 are used during actual training?
While having more points might seem conducive to learning better temporal and spatial codes, I'm curious about the rationale behind utilizing a relatively smaller quantity.