Hello, I encountered two problems when using find_object_3d

[Kane1440]

Hello, I encountered two problems when using find_object_3d
(1) Small objects can be identified but their depth information cannot be detected, so their spatial position cannot be obtained.
(2) In rviz, it is observed that the object's position relative to the camera is incorrectly oriented, because I placed the object between the camera and the robotic arm, but the object actually rotated 90°. What adjustments should I make?
I'd be grateful if you could help.

[matlabbe]

(1), if no depth can be detected, another approach would have to be used to know the 3D pose, like feeding height of the object.
(2), the detection would be in camera frame, maybe there is a missing optical rotation from your TF tree?