how to obtain 3d bounding box of the object from its {s,t,R}

autonomousvision / giraffe

This repository contains the code for the CVPR 2021 paper "GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fields"

https://m-niemeyer.github.io/project-pages/giraffe/index.html

MIT License

1.24k stars 160 forks source link

how to obtain 3d bounding box of the object from its {s,t,R} #51

Open PeizeSun opened 2 years ago

PeizeSun commented 2 years ago

Thanks for your excellent work ! I am really surprised by the controllable object generation !

I wonder if there is a way to extract 3d bounding box of the object from its affine transformation {s,t,R}.

I tried to establish the equation between the 3d bounding box and {s,t,R}, but failed. I transform a cube by equation(6) in the paper, but it seems their relationships do not follow equation(6). Do you have any suggestion ?

SongQiPing commented 2 years ago

basically I know what you mean, but practically in the model, there is no true bounding box of objects. I think you can draw a unit cube in object coordinates, and project it into the image.