GARfVDB: Segmentation cluster visualization

[swahtz]

Run with
python visualize_segmentation_clusters.py -s garfvdb_logs/run_2026-01-30-16-14-51/checkpoints/00036608/train_ckpt.pt -r out.ply --scale 0.2

[Copilot]

Pull request overview

This PR adds functionality to visualize segmentation clusters in GARfVDB by introducing a new visualization script that performs dimensionality reduction (PCA + UMAP) and clustering (HDBSCAN) on Gaussian splat features, then displays the results in an interactive 3D viewer.

Changes:

• New visualization script for segmentation cluster analysis using cuML-accelerated algorithms
• Enhanced view_checkpoint.py to properly handle world_to_camera transformations
• Added cuML, cuPy, and matplotlib dependencies for GPU-accelerated clustering
• Refactored model.py to extract common encoder grid sampling logic and add per-Gaussian affinity feature extraction

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 10 comments.

Show a summary per file

File	Description
visualize_segmentation_clusters.py	New script for clustering and visualizing segmentation features with PCA/UMAP reduction and HDBSCAN clustering
view_checkpoint.py	Added world_to_camera parameter to render_segmentation_image method and reordered imports
pyproject.toml	Added cuml dependency to Python package requirements
garfvdb_environment.yml	Added cuml, cupy, and matplotlib conda dependencies
garfvdb/model.py	Extracted _sample_encoder_grids_at_gaussians helper method and added get_gaussian_affinity_output for per-Gaussian feature extraction

---

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