Fixes returned tensor for normals in TiledCamera

[Rabbit-Hu]

Description

Fix a bug with TiledCamera. The 4-th channel in the "normals" annotator output should be discarded before reshaping into [B, H, W, 3]. Please find more details in the issue below:

Fixes #4239

Type of change

• Bug fix (non-breaking change which fixes an issue)

Screenshots

Before	After

Checklist

• I have read and understood the contribution guidelines
• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have updated the changelog and the corresponding version in the extension's config/extension.toml file
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[Rabbit-Hu]

Hi reviewers! I wonder if I need to (and how to) create tests and/or update the changelog for this bug fix.

[Mayankm96]

Thank you for the fix! Could we also modify the unit tests to capture this case?

[Rabbit-Hu]

Hi @Mayankm96 , thank you for your reply! I am happy to update the unit tests, but unfortunately I have run into issues running the unit tests for TiledCamera (#4255).

[Rabbit-Hu]

Hi @Mayankm96 ! I have updated the unit test to capture the previous bug.

I added an assertion to verify that the norms of the normals should be approximately 1. The previous (buggy) version would fail here because of the shift.

I also updated several existing assertions for the camera data keys in other test cases for TiledCamera, as my fix added a new key normals_padded to the data dict.

I have verified that the fixed code can pass all test cases in the updated unit tests in test_tiled_camera.py.

[ooctipus]

If I understand the issue correctly, the fix is discarding the fourth channel before reshaping. But I see implementation added a new output_type called normal_padded, is that needed?

[Rabbit-Hu]

Hi @ooctipus , thank you.

I just realized that adding a new output key was indeed unnecessary. I changed it to a simpler fix.

The unit test has been updated accordingly, capturing the previous bug without making unnecessary changes.