nx_compose.nx
buildroot/runtime/nx_compose.nx
about
nx_compose.nx -- compositional primitives (V2 of vision rollout).
Pure-math aesthetic measures that work on image PROPERTIES, never
on subjects. Per CAPTAIN_MORONI_DOCTRINE these are objective
scores; they never rank persons or assign beauty to humans.
Ships:
- nx_compose_thirds_score: rule-of-thirds alignment
- nx_compose_h_symmetry: left/right mirror symmetry
- nx_compose_v_symmetry: top/bottom mirror symmetry
- nx_compose_visual_centroid: weighted center-of-intensity
- nx_compose_balance_score: distance from geometric center
All scores in Q10 fixed-point (1024 = "perfect" depending on metric).
What this unlocks:
+ objective composition feedback ("subject not on thirds")
+ symmetry detection (architecture, portrait framing, reflections)
+ balance assessment (heavy on one side vs centered)
+ aesthetic measurement at the image level (never the subject)
genealogy_id: birkhoff_1933_aesthetic + datta_2006_56features
+ cropping_aesthetics_research
lineage_id: image_property_measurement (not subject ranking)
dependencies 2 imports · 4 importers
imports: syscalls.nxnx_image.nx
imported by: nx_aesthetics.nxnx_compose_test.nxnx_symmetry_group.nxnx_vision_pipeline_test.nx
structs
| none |
consts
| 35 | const NX_COMP_Q: i64 = 1024 |
functions
| 48 | func nx_compose_thirds_score(gray: *Image) -> i64 |
| 105 | func nx_compose_h_symmetry(gray: *Image) -> i64 |
| 132 | func nx_compose_v_symmetry(gray: *Image) -> i64 |
| 165 | func nx_compose_visual_centroid(gray: *Image, out_cx_q: *i64, |
| 199 | func nx_compose_balance_score(gray: *Image) -> i64 |