search

LiyaoTang / contrastBoundary

Contrastive Boundary Learning for Point Cloud Segmentation (CVPR2022)
MIT License
140 stars 11 forks source link

What is the offset in the input of the model? #27

Closed shoushouwang closed 1 year ago

shoushouwang commented 1 year ago

Hi, could you tell me what offset is in the input of the model？ Hope to receive your reply soon. Thanks!

LiyaoTang commented 1 year ago

Hi, thanks for your interest.

Could you provide more detail in your question? E.g. referring to the line of the code.

shoushouwang commented 1 year ago

Hi, thanks for your reply. I mean the word "offset" in line 98 of the pointtransformer_seg.py file. What is the data in "Offset" refers to？

LiyaoTang commented 1 year ago

Hi,

Since we have multiple point clouds of different sizes in the same batch, we stack them along the 1st axis and thus need to remember the index that starts a new point cloud.

The offset is precisely for this use. You could also check the original point-transformer repo for a better understanding :)