Abstract: Point cloud classifiers have been recently demonstrated to be vulnerable to backdoor attacks. The infected model functions normally on clean data, yet its predictions are errors when ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results