CloudAnalyzer / Public Demo

Public LiDAR map inspection, browser-side, no custom backend.

This Pages build publishes a static ca web bundle generated from the public hdl_localization sample map. The demo keeps progressive loading, point picking, heatmap thresholding, and trajectory inspection active while serving only HTML and JSON.

Open Live Demo Compare Clouds 3DGS Evaluation SLAM Leaderboard Open Repository

What This Shows

  • Progressive chunk loading on a static host.
  • Heatmap thresholding without a Python server.
  • Point picking with XYZ and nearest-reference distance.
  • Trajectory overlay and linked error timeline inspection.

Browser Compare

Drag & drop two PLY / PCD files for an instant heatmap and approximate Chamfer distance — no install, no upload. For CI-grade AUC / F1 gates, use ca evaluate locally.

Compare Point Clouds

3DGS Rendered Evaluation

Render a Gaussian Splat at nerfstudio camera poses, score PSNR / SSIM / LPIPS against reference views, and optionally combine with Chamfer / AUC geometry QA — ca rendered-evaluate in one command.

View 3DGS Demo

SLAM Leaderboard

Frozen regression snapshots for KISS-ICP, KISS-SLAM, and small-gicp on bundled synthetic suites. Each row links to the benchmark JSON, HTML report, and a static ca web-export heatmap viewer.

View SLAM Leaderboard

Perception Batch Demo

Batch comparison of a geometry-first non-deep baseline and a higher-fidelity deep baseline on the same public RELLIS-3D LiDAR frame. Demonstrates ca batch --evaluate with AUC, Chamfer, inspection commands, and quality-gate status.

View Perception Report

Detection & Tracking Demo

3D object detection (mAP/AP) and multi-object tracking (MOTA/ID switches) evaluation on public RELLIS-3D-derived examples. Compares good and regressed candidates with quality gates.

View Detection & Tracking Report

Build It Locally

After installing the package, rebuild the public demo bundles with:

python scripts/build_public_demo.py --output docs/demo/hdl-localization-map python scripts/build_perception_demo.py --output docs/demo/perception python scripts/build_detection_tracking_demo.py --output docs/demo/detection-tracking

Data Note

Each demo ships with an attribution file generated next to the bundle. The map viewer uses the BSD-2-Clause `hdl_localization` sample map, while the checked-in perception report is derived from the public RELLIS-3D example bundle and carries its CC BY-NC-SA 3.0 attribution note. The perception candidates are deterministic demo baselines generated from the same public frame. The map viewer trajectories are synthetic demo trajectories generated by CloudAnalyzer.