Skip to content

ROADMAP.md

Latest commit

 

History

History
74 lines (59 loc) · 4.2 KB

ROADMAP.md

File metadata and controls

74 lines (59 loc) · 4.2 KB
Raw

ThreeCrate Roadmap

This roadmap is written the same way our benchmarks are: honestly. It says plainly where ThreeCrate already wins, where it still trails Open3D/PCL, and exactly what work closes each gap. Every item links to a tracking issue — most are self-contained and a great way to start contributing.

New here? Look for good first issue and help wanted.

Where we stand today

Measured on full-resolution TUM RGB-D, KITTI, and nuScenes-mini frames against Open3D 0.19 (CPU, same machine). See docs/benchmarks.md for the full tables and reproduction command.

Workload Status vs Open3D
File read (raw float parse) ✅ Ahead 1.8x–2.2x faster
Voxel downsampling (centroid) ✅ Ahead 1.6x–1.8x faster
Normal estimation ⚠️ Behind at scale 0.57x–1.09x
Single-scale ICP ⚠️ Behind at scale 0.71x–0.99x
PCL comparison ⏳ Not yet measured —

Near-term: close the honest gaps

These are the concrete, measurable items that move the benchmark and the credibility story. In rough priority order:

  • Flat-layout kd-tree — done (#176). The pointer/Box tree is now a contiguous, index-referenced Vec<KdNode>. k-NN results are identical (all 201 algorithm tests pass); a same-machine A/B measured a consistent ~8–10% speedup on normal estimation and ~5–9% on ICP. It does not close the Open3D gap on its own — normals are still ~0.5x on large clouds — because the dominant remaining cost is per-point PCA and single-threaded correspondence search, not tree layout. That work continues in #177.
  • Dense ICP on large clouds — at parity on small clouds, still ~0.7x on KITTI/TUM even after the flat kd-tree. Parallelising correspondence search and per-point PCA (rayon) is the next lever. → #177
  • Integrate PCL into the benchmark table — the PCL harness is written and builds (scripts/pcl_bench/); it just needs to be run in a shared environment and folded into the published numbers. → #179
  • Realistic ICP target + accuracy comparison — today's benchmark tests per-iteration speed against a near-identity transform, not registration accuracy. → #180 (good first issue)

Medium-term

  • Competitive GPU compute — cache pipelines, async readbacks, and a GPU-resident spatial index so GPU knn/normals/icp beat CPU (voxel and TSDF already do). → #178
  • Fix the GPU TSDF buffer-cast panic — done (#175). The readback cast a mapped GPU buffer (8-byte aligned) straight into repr(align(16)) structs; now it copies into a correctly aligned Vec. All TSDF tests pass, no #[ignore].
  • Broader format coverage and streaming improvements across threecrate-io.
  • Python API parity with the Rust surface (threecrate-python).

Longer-term / exploratory

  • WebAssembly target for in-browser point-cloud processing.
  • More global-registration and segmentation algorithms.
  • Realistic, published accuracy benchmarks (not just speed) across libraries.

How to help

  1. Pick an issue above (or any good first issue).
  2. Read CONTRIBUTING.md for setup and guidelines.
  3. For perf work, include before/after benchmark numbers — the reproduction command is in docs/benchmarks.md.
  4. Open a draft PR early; we'd rather help shape it than review it cold.

Have an idea that isn't here? Open a discussion or an issue.