The obtained vectors are transferrable, sample-independent and preserve shape smoothness and occlusions.
Our approach constrains 3D points to slide along their sensor view rays while neither adding nor removing any of them.
We achieve this with 3D-VField: a novel method that plausibly deforms objects via vectors learned in an adversarial fashion. In this work, we substantially improve the generalization of 3D object detectors to out-of-domain data by taking into account deformed point clouds during training. As 3D object detection on point clouds relies on the geometrical relationships between the points, non-standard object shapes can hinder a method’s detection capability However, in safety-critical settings, robustness on out-ofdistribution and long-tail samples is fundamental to circumvent dangerous issues, such as the misdetection of damaged or rare cars.
