Geometric Benchmark Demo

Status: Demo-only benchmark walkthrough Device: CPU-compatible

This walkthrough keeps the historical geometric benchmark example, but the public claim is now narrow: the shipped pair uses retained synthetic ShapeNet-style data and demo-only benchmark wiring to illustrate point-cloud evaluation. It is not the canonical benchmark-grade ShapeNet runtime.

Files

• Python script: geometric_benchmark_demo.py
• Jupyter notebook: geometric_benchmark_demo.ipynb

Run It

python examples/generative_models/geometric/geometric_benchmark_demo.py
jupyter lab examples/generative_models/geometric/geometric_benchmark_demo.ipynb

What This Demo Actually Uses

• ShapeNetDataset with data_source="synthetic" and explicit demo-mode metadata
• PointCloudGenerationBenchmark through typed benchmark configuration objects
• PointCloudMetrics and PointCloudModel for retained point-cloud evaluation
• training and plotting helpers around a synthetic airplane-style subset

Why It Is Demo-Only

• benchmark-grade ShapeNet assets are not auto-downloaded by the supported runtime
• the shipped walkthrough intentionally opts into retained synthetic data for reproducibility
• the public CLI does not ship a supported benchmark runner for this workflow

Use This When

Use this pair when you want to inspect the retained geometric benchmark API shape, run a synthetic point-cloud demo, or adapt the code to your own benchmark-grade assets from Python.