feature op• Data kinds: voxel → flow_dense
• Call: import fullseye as fs; fs.ledger.edt_jfa_vector(seed_bool, device='cpu') (to call the implementation directly, import match3d; match3d.edt_jfa_vector(seed_bool, device='cpu'); from the registry, ops3d.get("edt_jfa_vector"))
• Return value through the ledger: fullseye.ledger.edt_jfa_vector(...) returns **only the declared out type flow_dense** (the underlying function also returns auxiliary values). When you need what was dropped, use fullseye.ledger.edt_jfa_vector.raw(...) or call match3d.edt_jfa_vector directly.
• GPU: this op has a GPU path (device="cuda")
> This operator's description has not been translated yet. The original text follows as it is.
各 voxel から最近 seed への変位 `(3, D, H, W)(flow_dense`、dz, dy, dx [voxel])を GPU の JFA で。
`edt_jfa` は距離の値だけを返すが、JFA は内部で最近 seed の座標を運んでいる —— それをそのまま出す(追加コストほぼ 0)。
CPU / scipy 経路は `vol_nearest_seed_vector`(同じ値、N≤160 で厳密一致を実測)。seed が無ければ ValueError
(`edt_jfa` は 1e6 に飽和させるが、向きに「無限遠」は無い)。返りは torch float32(台帳経由では numpy)。
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• nearest_seed_partition — py -3.11 examples_3d/nearest_seed_partition.py
flow_dense as input)feature)sobel3d · hessian3d · curvature_maps · edt_jfa · vol_frangi · vol_local_std · vol_local_thickness · vol_orientation_coherence
*Provenance: match3d.py — 3D operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.