amine-chaabouni

You should use `downpcd = pcd.voxel_down_sample(voxel_size=downsample_voxel_size) ` instead of the proposed solution

Hi, Thank you for your response. So, training the pre-trained model on a similar dataset as the point clouds I'm using will make it performant for my case?

It can be downloaded, you're right, but the files are somehow corrupted and I can't extract them. I tried with 3 different computers on different connections, with different OS.

I just figured out that you're using part of DeepI2P's script And I'm guessing that you mean to use `CorrI2P` here instead of `DenseI2P` https://github.com/rsy6318/CorrI2P/blob/963724b5198ba0caa8236b640c76751af635d854/eval_all.py#L5

You said here that you preprocessing is different from DeepI2P https://github.com/rsy6318/CorrI2P/issues/2#issuecomment-1236080480 However, it looks exactly alike. Can you explain the difference?