NVlabs / FoundationPose

[CVPR 2024 Highlight] FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects
https://nvlabs.github.io/FoundationPose/
Other
955 stars 99 forks source link

run_nerf.py error "[WARNING] batch has 0 intersections!!" #90

Closed Celinna closed 2 weeks ago

Celinna commented 3 weeks ago

Hello,

Thank you for the great work.

I am trying to generate the textured mesh and .obj file of a custom object with 16 reference images. I collected rgb, depth (aligned), segmentation masks, and pose (object wrt. camera) data. I get "[WARNING] batch has 0 intersections!!" when running run_nerf.py on this data.

I suspect the object poses are not in the correct form. Could you clarify how to get the data in the cam_to_ob folder for model-free implementations?

Thanks

wenbowen123 commented 3 weeks ago

you may need to get the inverse pose, see https://github.com/NVlabs/FoundationPose/issues/49#issuecomment-2068038921

Celinna commented 2 weeks ago

Hello,

Thank you your suggestion worked!