search

lucidrains / perceiver-pytorch

Implementation of Perceiver, General Perception with Iterative Attention, in Pytorch
MIT License
1.1k stars 134 forks source link

How to use this model for point cloud? #39

Open sangyx opened 3 years ago

sangyx commented 3 years ago

Hi, I'm trying to use this model for point cloud task. However, the shape of my input data is (B, N, C), where B is the batch size, N is the number of points, C is the dim of feature. Which params I need to modify to apply this model to my task?

Forbu commented 3 years ago

You can try : model = Perceiver( input_channels = C, # number of channels for each token of the input input_axis = 1, etc)

It will suppose you have only one axis (here all of your points) and you have C channels like the dimension of the features.

dnhkng commented 3 years ago

I think the point's location also needs to be converted to Fourier Features?