Hi, thanks for the incredible framework for point cloud processing.
However, when I read your code on the completion task in examples/completion.py, I noticed that the input of the network is still the complete point cloud:
In line 541,
sin = ME.SparseTensor( features=in_feat, coordinates=data_dict["coords"], device=device, )
where data_dict['coords'] refers to point cloud which is not randomly cropped.
Hi, thanks for the incredible framework for point cloud processing. However, when I read your code on the completion task in examples/completion.py, I noticed that the input of the network is still the complete point cloud:
In line 541,
sin = ME.SparseTensor( features=in_feat, coordinates=data_dict["coords"], device=device, )
where data_dict['coords'] refers to point cloud which is not randomly cropped.I wonder if I have misunderstood the code.