Lower live Leap packed stereo to depth socket

This commit is contained in:
Meta Vens 2026-05-30 21:54:58 +01:00
commit 2c197cb3ae
9 changed files with 247 additions and 36 deletions

View file

@ -41,19 +41,21 @@ The research ledger is
That note is not a dependency grant: CUDA implementations, TensorRT ports, and
monocular model demos are provenance only. The live owner is Fensalir D3D12
compute over Mimir-declared texture resources and calibration state.
The first contract slice now exists in `MimirStereoDepthConfigurations` and
`MimirFensalirFieldLowering.BuildStereoDepthCandidateFrame`: a calibrated
stereo pair profile references caller-declared shader-readable left/right input
textures, emits a GPU-resident compute-writable R16F disparity `SurfacePage`,
an R8 confidence texture, a Height FieldEvidence claim planned to the
`SurfacePage` backend, and an `AquariumFieldStereoDepthLowering` sidecar tying
profile, calibration, camera pair, inputs, disparity output, disparity
settings, SGM constants, and depth range together. The libSGM-provenance profile
currently publishes min disparity 0, 128 disparities, four aggregation paths,
census radius 2, P1 8, and P2 96. This is the socket for the kernel, not the
kernel. Fensalir's renderer now reports whether those lowerings are
dispatch-ready after planning/resource resolution, without writing placeholder
depth values.
The first live Leap slice now exists in `MimirStereoDepthConfigurations` and
`MimirFensalirFieldLowering`. The generic
`BuildStereoDepthCandidateFrame` still lowers caller-declared rectified
left/right textures into a GPU-resident compute-writable R16F disparity
`SurfacePage`, an R8 confidence texture, a Height FieldEvidence claim, and an
`AquariumFieldStereoDepthLowering` sidecar. The live
`BuildLeapPackedStereoDepthCandidateFrame` path now recognizes the current
`LeapStereoIr` rolling video window, reuses its declared packed R8G8 texture as
both left and right input resources, names the R/G lane observations, and emits
the same Height/depth lowering automatically from the runtime frame. The
calibration id is explicitly pending, so the output is disparity evidence, not
metric geometry truth. The libSGM-provenance profile currently publishes min
disparity 0, 128 disparities, four aggregation paths, census radius 2, P1 8,
and P2 96. Fensalir has the first packed-Leap D3D12 compute kernel installed as
a crude SAD/block-match disparity writer; it is not full SGM yet.
Those claims are not inherently pixel-sized. Pixel-level resolve consumes the
reservoir, but claim support is chosen from the represented field. Smooth