camera based 3D bounding box detection (transformer multicue)

ApolloAuto / apollo

An open autonomous driving platform

Apache License 2.0

25.28k stars 9.72k forks source link

camera based 3D bounding box detection (transformer multicue) #7746

Open SuriyaNitt opened 5 years ago

SuriyaNitt commented 5 years ago

Hi,

I am curious to understand how the camera pipeline estimates 3D bounding boxes from 2D bounding boxes.

Is there a paper, that you guys can point to?

Is the 3D information out of this module used in fusion?

If so, why is not included in the demo record file?

Thanks in advance

curiousKid007 commented 5 years ago

I am wondering if you could share some more information about this issue. I am unable to understand how exactly do we calculate the center of the object and how do we use (H,W,L) and orientation to project 3D Bounding boxes.

Thanks