SCALE ISSUE

[Manu752]

Hi !

I’m using FoundationPose to detect a cube . My real object is 5 cm (0.05 m), but FoundationPose only works reliably if I scale the mesh ×10 in Blender (so only when the object is 0.5 m).

My camera intrinsics are correct.

The exported mesh dimensions are verified with trimesh (0.05 m bounding box).

FoundationPose inference works with the scaled mesh.

Can I simply divide the translation vector by 10 after inference (while keeping rotation untouched) ? or is there a recommended way to handle unit scaling in FoundationPose to avoid this mismatch ? any other way I can fix this error ?