v0.1.0 — QA Platform for Mapping, Localization, and Perception

Highlights

CloudAnalyzer v0.1.0 — a QA platform for mapping, localization, and perception point cloud outputs.

32 CLI commands across 8 groups

• Point cloud evaluation: Chamfer, F1, AUC, Hausdorff with quality gates
• Trajectory evaluation: ATE, RPE, drift, lateral/longitudinal decomposition, coverage
• Ground segmentation QA: Precision, recall, F1, IoU via voxel-based comparison
• Config-driven quality gates: cloudanalyzer.yaml for unified CI/CD QA
• Baseline evolution: Promote / keep / reject decisions with history management
• Processing: Downsample, filter, split (with metadata.yaml), merge, convert
• Browser inspection: ca web / ca web-export with heatmap, trajectory, point picking
• Reports: Markdown and HTML with trajectory overlay and error timeline

7 experiment-driven slices

Each core feature was developed by comparing 3 competing implementations, selecting the best, and stabilizing the minimal contract.

CI integration

• config-quality-gate.yml: Run ca check as a reusable GitHub Actions workflow
• baseline-gate.yml: Run check + baseline-save + baseline-decision in one job
• public-benchmark-pack.yml: Validate benchmark packs against Stanford Bunny data

Public demos

• Stanford Bunny Viewer
• Perception QA Dashboard

Stats

• 389+ tests passing
• 87 source files, mypy clean
• Python 3.10+, MIT License