PETR
PETR copied to clipboard
a question of this permutation
https://github.com/megvii-research/PETR/blob/2f55c1597e39c4d89dfd8ef07df29a955299d5fc/projects/mmdet3d_plugin/models/dense_heads/petrv2_dnhead.py#L370 why there is (W, H, D) not (H, W, D) ?
For coordinate transformation, (x, y, d) is used in image space, so we should put W in front.