Could the pixel-level contrastive learning be adapted to 3D classification or segmentation task？

[Lazzzycoding]

Dear authors @lucidrains @cjrd :

Thanks for your innovative works!

I wonder to know whether this pixel-level contrastive learning work be adapted to 3D classification task or segmentation task? Which component is the most effective one in the current pixel-level contrastive learning framework? Thanks a lot.

Best wishes