Andreas K

In PointNet they used T-Nets to achieve invariance under geometric transformations, however in PointNet++ the T-Nets are not used anywhere. Does the PointNet++ use something else to achieve it?

Update correspondences.py

1

``` closest_points = np.mean(closest_points, 1, keepdims=False) ``` I think it can be replaced by `closest_points.squeeze()`, since the dimension of `closest_points` is (N, 1, 3) and by taking `np.mean` along the...

Are the MLPs same as in the paper?

1

In the paper, the first MLP (64, 64) has two layers and the second MLP (64,128,1024) has three layers. In the implementation of PointNetfeat [here](https://github.com/fxia22/pointnet.pytorch/blob/f0c2430b0b1529e3f76fb5d6cd6ca14be763d975/pointnet/model.py#L87), the first MLP has just...