Fix non-finite depth display normalization

[sylvesterkaczmarek]

Description

normalize_camera_output_for_display() normalized depth images using the raw tensor maximum. Depth camera outputs may contain inf for no-hit pixels and can contain other non-finite values, so the maximum became non-finite; dividing by it could produce NaN pixels and suppress finite depth contrast.

This change zeroes non-finite depth values before computing the display scale, preserving finite depth normalization while keeping no-hit pixels black.

Validation

Adds unit coverage for:

• mixed finite, inf, and NaN depth values across supported depth display types;
• all-non-finite depth input.

Type of change

• Bug fix

[greptile-apps]

Greptile Summary

Fixes depth-image display normalization by replacing non-finite values with zero before calculating the finite display scale.

• Preserves contrast among finite depth values while rendering no-hit pixels black.
• Adds coverage for mixed finite/non-finite and entirely non-finite depth tensors.
• Documents the correction in the changelog.

Confidence Score: 5/5

The PR appears safe to merge with no actionable correctness, security, or integration issues identified.

The implementation sanitizes non-finite values without changing tensor shape or device, computes normalization from finite depths, and handles entirely non-finite input without division.

Important Files Changed

Filename	Overview
source/isaaclab/isaaclab/utils/images.py	Sanitizes non-finite depth values before maximum-based normalization, correctly preventing non-finite scales and outputs.
source/isaaclab/test/utils/test_images_display_nonfinite.py	Covers every supported depth display type with mixed finite/non-finite input and verifies all-non-finite input produces zeros.
source/isaaclab/changelog.d/sylvesterkaczmarek-fix-depth-display-nonfinite.rst	Accurately documents the corrected depth-display behavior.

_{Reviews (1): Last reviewed commit: "docs: note depth display normalization f..." | Re-trigger Greptile}

[isaaclab-review-bot]

Isaac Lab Review Bot

The patch narrowly updates depth display normalization to replace non-finite values with zero before selecting the display scale, with focused regression coverage and a package changelog fragment.

• Design and architecture: The change remains within the display-normalization utility and preserves the existing convention that invalid or no-hit depth pixels display as black. It does not shift responsibility to camera producers or downstream callers.
• API: The function signature, output shape, float32 conversion, and device placement are unchanged. Non-finite depth values now deterministically produce zero-valued display pixels. The added torch.where also ensures a fresh depth output tensor, but aliasing is not documented as part of the API contract.
• Implementation: The revised depth path removes NaN and positive or negative infinity before computing the maximum, preserving contrast among finite positive depths. The existing max-value guard correctly leaves all-non-finite input as zeros. Tests cover mixed finite/non-finite values for all three supported depth display types and the all-non-finite boundary case; the required source-package changelog fragment is present.

No blocking issues. No inline issue met the actionable-evidence threshold; the assessment above records the review feedback.

Automated review; human maintainers own approval decisions.

[AntoineRichard]

Thanks a lot for the PR, and contributing to IsaacLab! @kellyguo11 @hujc7 can I have another review?