Collate_functions Args: batch (list): input batch for training

Project a pointcloud into a BEV picture

inputs : x : input feature maps( B X C X W X H) returns : out : self attention value + input

inputs : x : input feature maps( B X C X N) returns : out : attention weight + input feature(

Downsample pointcloud to 4096 points Modify based on the version from https://blog.csdn.net/SJTUzhou/article/details/122927787

:param i, j: index of pixel in equirectangular :param rotation: (yaw, pitch, roll) in degree :return: rotated index of pixel

:param image: (H, W, C) :return: rotated (H, W, C)