visual of the pseudo boxes

Question

visual of the pseudo boxes

xuxiaoxxxx opened this issue a year ago · comments

Thanks for your work in 3D detection field! After I visual the pseudo boxes, I find the result is very worse. Do I need to multiply by the rotation matrix?

Dawn · Answer 1 · Mon Sep 04 2023 16:26:44 GMT+0800 (China Standard Time)

Then I realised that the dataset you provided is not divided into train and val like the default division of the ScanNet dataset, can you tell me the basis of the division?

Dawn · Answer 2 · Mon Sep 04 2023 16:29:01 GMT+0800 (China Standard Time)

Since I found that basically all the scenarios in the val dataset were present in the train dataset, does that make the val dataset invalid for evaluation? I.e. a complete overfitting situation