3D object Pose Estimation with Pose CNN decoder taking too long to run

@rkjha1 On the inference issues, can you show us:

  • The O.Detection inference (BBox)? Is it working fine?
  • The Segmentation Decoder output (TensorViewer) you get from the network?
  • Did you adjust object_T_box_center and box_dimensions parameters of the Detections3Viewer node?

I don’t get any error when launching the Unity. I get ‘Segmentation fault (core dumped)’ when I play the scene and Unity crashes. This occurs only when I implemented the randomization of object as shown in https://drive.google.com/file/d/194qsZqkEEZgxUa7Ld2bLZt1ykt7HQ4ji/view.

  1. Driver version: 450.80.02, GPU: RTX 2060 Super
  2. I quite didn’t understand the question.
  3. The default scene which uses Scenario Manager works fine. I even generated dataset for dolly, small and even the duplo object with that method.
  4. No, this is not observed in other scenes.

The scene without Scenario manager where it crashes is shown here: https://drive.google.com/file/d/1tp9BMJ8x8E9PEdg88i8N-kJaJ2tydYg_/view?usp=sharing

The default scene with which it works in shown in this video: https://drive.google.com/file/d/1PMztWQfGAhycEyOnPl7lgH1taSytjDzh/view?usp=sharing

As the default settings work, I’m thinking if there is any mistake I did in the modifications. Please look at the attached video above and let me know.