Improve Segmentation, Logging, and Dataset Conversion

[sokovninn]

Purpose

This PR introduces multiple improvements and fixes to enhance annotation processing, logging, and dataset conversion.

Specification

• Fixes:
  • Skip num_objects_range check when no generation is involved.
  • Correct bounding box visualization image ordering.
  • Skip empty segmentation annotations during conversion to coco.
• Features:
  • Implement segmentation mask smoothing (2x-10x smaller annotation file size).
  • Integrate loguru for enhanced logging capabilities.
  • Update Luxonis dataset converter with latest improvements.
• Refactor:
  • Rename confidence threshold variable for better readability.

Dependencies & Potential Impact

• No known breaking changes.

Deployment Plan

1. Deploy changes to a staging environment for validation.
2. Monitor possible issues.
3. Roll out to production.

Testing & Validation

• Visual validation of segmentation mask smoothing.
• Logging behavior verified.
• Dataset conversion tested with sample Luxonis datasets.

[github-actions]

Test Results

  6 files    6 suites   3h 31m 36s ⏱️
 79 tests  58 ✅  20 💤 1 ❌
474 runs  344 ✅ 124 💤 6 ❌

For more details on these failures, see this check.

Results for commit c7c2fea.

♻️ This comment has been updated with latest results.

[HonzaCuhel]

LGTM! 🚀🚀🚀