hi，can banmo be used for 3D reconstruction of cars？

[yushengjiexy]

hi，can banmo be used for 3D reconstruction of cars？

[gengshan-y]

The input viewpoint/pose is defined as
P_camera = R P_canonical + T
where (R T) is the rigid transformation that transforms 3D points in the object canonical coordinates to the camera coordinates. We use opencv coordinate systems (x-right, y-down, z-forward). To convert from opengl format, you want to take the negative value of y and z coordinates.

Let me know if there are other questions.

[gengshan-y]

Hi, BANMo can be applied to cars and other categories if rough camera viewpoints are given.
There are pre-trained models to estimate camera viewpoint/pose for cars, e.g., OcclusionNet / EgoNet (I don't have experience running those).
Once you have the camera poses, please follow this to set up the config files.

[yushengjiexy]

Thanks for your reply！
When you say rough camera viewpoints, do you mean the camera's own position or the car's position in the camera's view？

[yushengjiexy]

sorry, I am a beginner, my question is more basic.
Can you explain what the object canonical coordinate is, or provide me with some papers? Is the object canonical coordinate similar to the world coordinate, or is the object coordinate fixed on the object moving with the object?
Thank you very much!

[gengshan-y]

I see, this might help -- by canonical coordinate, I mean the model space. It's the coordinate system attached on the object.

[yushengjiexy]

Thank you！
I have no further questions for the time being. I'm going to do it now. Thank you for your guidance.