Inference script

[muhabdulhaq]

Hi, thanks for sharing this great work!

I’m trying to use 360-DFPE and I have a couple of questions:

• The repo mainly explains the evaluation process, but I couldn’t find how to run inference on new 360° images. Is there an inference script or recommended steps to do that?
• In the dataset, there is a label.json file. Was this label generated manually, or is it automatically produced by 360-DFPE or another tool?

Thanks a lot!

[EnriqueSolarte]

Hi,

Thank you for your interest in our project.

[Q] Is there an inference script or recommended steps to do that?
[A] Yes, you want to check the section Computing metadata in our README.md.
This section aims to register camera poses with estimated layouts by using the script main_compute_metadata.py.
The camera poses and the layouts were computed by using OpenVSLAM and HorizonNet.

[Q] Was this label generated manually, or is it automatically produced by 360-DFPE or another tool?
[A] We label this data manually using the depth information in the data (as top-down view of the point cloud scene)

Best regards