how can i train xyz points without normal?

[azizanajeeb]

when i train my dataset without normal i get and error:
RuntimeError: Given groups=1, weight of size 64 6 1 1, expected input[32, 3, 32, 512] to have 6 channels, but got 3 channels instead
is there a way i could resolve this?

[geyang]

Take a look at the source.

@DylanWusee there is a bug in the released code. Signature of the classifier is wrong.

[chenyanjie519]

我也遇到了这个问题，请问您解决了吗

[chenyanjie519]

Where can I see the source ？

[heexu976]

I have the same problem

[294coder]

收到

[dhliubj]

I encountered the same problem, reference (https://github.com/yanx27/Pointnet_Pointnet2_pytorch/) to the corresponding code to make the following changes, you can solve the problem of this channel.

Modify line 16 of the model/pointconv.py file:
self.sa1 = PointConvDensitySetAbstraction(npoint=512, nsample=32, in_channel=feature_dim+3, mlp=[64, 64, 128], bandwidth = 0.1, group_all=False)

Modify in_channel=feature_dim+3 to in_channel=feature_dim. Then you can successfully process the data without normal vector.

@heexu976 @azizanajeeb @chenyanjie519