For paper "PointCloud Saliency Maps" (To appear in ICCV 2019 oral presentation, 187 out of 4303)

In the folder "point-master", we provide the code for testing our method on PointNet. The code is written based on the code from (PointNet https://github.com/charlesq34/pointnet)

Please first refer to https://github.com/charlesq34/pointnet to find how to train and evaluate a ModelNet40 model. Afterwards, run saliency.py to attack the model.

If there is any question, pls contact tzheng4@buffalo.edu