Pointnet_Pointnet2_pytorch

Pointnet_Pointnet2_pytorch copied to clipboard

what is data augmentation trick you added? I want to know the classification accuracy without the additional augmentation trick. what is the classification accuracy with pure augmentations which are mentioned in pointnet++ official paper?

it seems to be adding normal vx,vy,vz to the points' x,y,z data

@dogyoonlee and @ashlet It seems like they are not doing rotation and jitter but only doing random point drop out, scaling, and shifting.

@dogyoonlee and @ashlet It seems like they are not doing rotation and jitter but only doing random point drop out, scaling, and shifting.

I agree with you and do you think this affects the result to a considerable degree? By the way, I'm confused that how can we define the training set, validation set, and test set in this implementation. Is validation set the test set?