Add beginner-friendly LiDAR segmentation tutorial (RELLIS-3D compatible, no heavy dependencies)

[Khiladi-786]

Overview

This PR adds a beginner-friendly LiDAR segmentation tutorial notebook to the examples/ directory.

Unlike existing approaches, this tutorial is designed to be:

• Lightweight (no heavy dataset/model download required)
• Beginner-friendly for new contributors
• Focused on understanding PerceptionMetrics pipeline

---

Key Features

• Step-by-step explanation of LiDAR segmentation workflow
• Synthetic data fallback (runs without RELLIS-3D dataset)
• Manual IoU & mIoU computation for learning purposes
• Clear visualization of segmentation outputs

---

Why this is useful

• Helps new contributors understand evaluation pipeline
• Removes dependency on large datasets (common blocker)
• Complements existing tutorials with a simpler entry point

---

Added

• examples/tutorial_lidar_segmentation_rellis3d.ipynb

---

Tested

• Runs successfully without dataset/model
• Verified in clean Python environment

---

Notes

This tutorial is intentionally simplified to prioritize learning and accessibility. It can be extended later to include full dataset/model integration if needed.

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Khiladi-786]

Refined the optional RELLIS-3D section with clearer guidance and added references to the dataset loader.

The tutorial now:

• remains beginner-friendly (synthetic setup)
• supports extension to real dataset workflows

Happy to make any further adjustments if needed.

[Khiladi-786]

Added basic SemanticKITTI dataset support as discussed.
This extends the evaluation pipeline beyond synthetic data and aligns with real-world LiDAR benchmarks.

[Khiladi-786]

Happy to further extend this if needed — for example:

• adding evaluation example using SemanticKITTI
• improving dataset documentation
• aligning with existing dataset interface patterns

Let me know what direction would be most useful for the project.

[dpascualhe]

Closing since there is no associated issue (feel free to reopen after creting the corresponding issue). Also, adding SemanticKITTI will part of the GSoC project if possible and so it won't be done for now.

[Khiladi-786]

Thanks for the clarification!

Understood — I’ll create a proper issue first to align with the repository workflow and scope.

I’ll also separate the contributions to keep things focused (e.g., tutorial vs dataset support), and reopen accordingly.

Appreciate the guidance!