Fetching the paper…

PETR: Position Embedding Transformation for Multi-View 3D Object Detection · Around