Skip to content

docs(splat-native): D-SPLAT-4 ADR-024 adoption callout (symmetric to cesium-osm §11) #476

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
AdaWorldAPI merged 2 commits into from
Jun 5, 2026
Merged

docs(splat-native): D-SPLAT-4 ADR-024 adoption callout (symmetric to cesium-osm §11) #476

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
1cb2576
docs(splat-native): D-SPLAT-4 ADR-024 adoption callout (symmetric to …
claude Jun 5, 2026
7d09f75
docs(splat-native): address codex P2 on #476 (SH band notation ℓ=3 → …
claude Jun 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 10 additions & 1 deletion .claude/plans/splat-native-ultrasound-v1.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -152,7 +152,16 @@ The palette codec is currently `(centroid_palette: u8, edge_weight: u8, scent: u

**Compose-table size doubles** (from 256×256×1 B to 256×256×2 B) — still fits in 128 KB, still L1-resident.

**Tests:** palette compose against ground-truth Gaussians; SH-basis discrimination (cosine ≥ 0.95 against analytic SH on phantom data).
**ADR-024 adoption (palette256 + HHTL codec, OGAR #39).** D-SPLAT-4 is named in `OGAR/docs/ARCHITECTURAL-DECISIONS-2026-06-04.md § ADR-024 Consequences` as one of the two queued ADR-024 adopters (paired with D-OSM-2 — see `cesium-osm-substrate-v1.md §11` for the geographic-side adoption note). Mapping ADR-024's four-step adoption checklist onto D-SPLAT-4:

1. **Prefix.** FMA NiblePath + SH basis-id — the per-region spatial frame (FMA class identity within the atlas + SH basis index spanning the ℓ≤3 carrier).
2. **Palette domain.** Quantized SH coefficients (16 coefficients spanning ℓ≤3 — degree 0: 1, degree 1: 3, degree 2: 5, degree 3: 7 — totaling 16, matching the §3 D-SPLAT-1 carrier shape) clustered per FMA region. Per-region palette captures dominant tissue echo signatures (planar fascia / spherical organ surfaces / Doppler-flow vascular).
3. **ρ-vs-reference target ≥ 0.99**, matching the `lance-graph-arm-discovery` aerial-codebook empirical anchor (ρ = 0.9973 vs cosine). Reference: analytic SH evaluation at the same view direction (the same reference used in the D-SPLAT-4 test plan below). Report empirical ρ on the **first FMA-region palette build** (default region: femur cortical bone — smallest tractable corpus; cf. OGAR PR #30 §6 FMA-bones litmus).
4. **Decode = const-table lookup.** Per-region SH palette is runtime const-table; decode path is zero-allocation. Compile-time HHTL where the palette is shared across regions (e.g. global basis-ID lookup for the 256 SH signatures).

The 256-ceiling escape hatches from ADR-024 apply directly: per-region palettes are the cheapest answer (different FMA region, different 256 SH-signature entries); hierarchical palettes (coarser at OrganSystem level, finer per Bone leaf) mirror the SH ℓ=0/ℓ=1 vs ℓ=2/ℓ=3 split. Palette-64K is reserved for measured-cardinality escalation, not speculation.

**Tests:** palette compose against ground-truth Gaussians; SH-basis discrimination (cosine ≥ 0.95 against analytic SH on phantom data); **NEW: empirical ρ-vs-reference reported on first FMA-region palette build (target ≥ 0.99 per ADR-024 adoption contract)**; per-region palette cardinality distribution (mean / p95 / p99) at FMA region granularity.

**Gates on:** D-SPLAT-1.

Expand Down