TWO of the generated object_data.json files are different from the author's

[lcl-git-3d]

After running this code:python -m megapose.scripts.run_inference_on_example barbecue-sauce --run-inference
The TWO I generate is:"TWO": [0.7242290881878867, 0.5557210741091521, -0.3535530397018472, 0.2041240084548291], [0.08103131502866745, 0.05744335055351257, 0.3665071129798889].
This results in a final pose estimate that is less accurate than the authors'.I don't know what the problem is, please help me!

[monajalal]

Hi I am confused about same. I was trying to construct an object_data.json for my test object and then noticed how

[{"label": "barbecue-sauce", "TWO": [[0.5436763454963882, 0.622191933760276, -0.4349827125343375, 0.3578873048305161], [0.1073753759264946, 0.07318738847970963, 0.4580722153186798]]}]

is which is confusing. You mention we should have x_min, y_min, x_max, y_max then why do we have an array of size four and then an array of size 3? Also, there is only one barbeque sauce here. What is TWO referring to?

[lcl-git-3d]

Hi,I know what the reason is that the fine estimation model did not load successfully. As for what the TWO is, the author says:the estimated pose is noted TWO (the world coordinate frame correspond to the camera frame). It is composed of a quaternion and the 3D translation.