nn.Conv2 and nn.Linear in 3d points operation

vision
1

In PointNet++, suppose the input point data shape is [1, 3, 1024, 32] where the 1024 the centroids, and 32 is the neighbors and 3 is xyz coordinate with batch size as 1.

The input data is passed into nn.Conv2d(in_channels=3, out_channels=64, kernel_size=(1, 1))

Can we use nn.Linear instead of nn.Conv2d:
Change the input point data shape as [1 * 1024 * 32, 3], then pass the data into nn.Linear(3, 64)

I am wondering if these two have the same effect, since the conv2d kernel size is (1, 1)?

2

Hi Elijah!

Yes (although I’m not sure what the point would be).

Consider the following example (with fewer centroids and neighbors for
convenience):

>>> import torch
>>> print (torch.__version__)
2.2.1
>>>
>>> _ = torch.manual_seed (2024)
>>>
>>> conv = torch. nn.Conv2d (in_channels = 3, out_channels = 5, kernel_size = (1, 1))
>>>
>>> x = torch.randn (1, 3, 2, 4)
>>>
>>> resConv = conv (x)
>>>
>>> lin = torch.nn.Linear (3, 5)
>>> with torch.no_grad():
...     _ = lin.weight.copy_ (conv.weight.squeeze())
...     _ = lin.bias.copy_ (conv.bias)
...
>>> resLin = lin (x.permute (0, 2, 3, 1).flatten (0, 2)).reshape (1, 2, 4, 5).permute (0, 3, 1, 2)
>>>
>>> torch.allclose (resConv, resLin)
True

Best.

K. Frank

3

Thank you. This is what I need.