fullseye

seg_wiring_exposure — EMPROOF wiring op

• データ種: labels2d × table → table
• 呼び出し: import fullseye as fs; fs.ledger.seg_wiring_exposure(labels, synapses, spacing=None, background=0, p=0.5) (実装を直接呼ぶなら import segcompare; segcompare.seg_wiring_exposure(labels, synapses, spacing=None, background=0, p=0.5)、台帳から引くなら opsemproof.get("seg_wiring_exposure"))

使い方

How much wiring each object stakes on staying one piece — from synapse end counts alone.

Cutting an object whose s synapse ends (synapses as in :func:seg_synapse_partners) fall s1 / s - s1 on the two sides costs exactly (s / 2n) * H2(s1 / s) bits of ends-level split (:func:seg_wiring_variation), and a hidden merge of two neurons inside one object costs the same expression as merge. So what an object can cost is fixed by its end count before any geometry is known:

• bound = s / 2n — the supremum over all cuts (reached when the ends halve). The bounds sum to n_ends / 2n: exactly 1 bit when no end lies on background.
• worst = (s / 2n) * H2(floor(s / 2) / s) — the largest cost a cut can actually reach.
• expected = (s / 2n) * E[H2(K / s)], K ~ Binomial(s, p) — the cost of a cut that puts each end on the cut-off side independently with probability p (default 1/2, the cut through the object’s median voxel with ends spread evenly over it). This is the count-only prediction; the gap to the observed cost is the spatial clustering of the ends.
• density = (s / 2n) / (voxels / N) — the object’s share of ends over its share of the N voxels of the volume. For an edit, wiring bits / pixel bits = density × (H2 of the end fractions / H2 of the voxel fractions): merges have the second factor near 1 (ends follow voxels), so their ratio is the density; a cut’s second factor is H2(s1 / s) / 1 and is what the binomial predicts.

Objects with no end are not listed. Returns labels, voxels, ends, density, bound, worst, expected (aligned arrays), n synapses, n_ends kept, n_background ends dropped, total_bound, total_expected.

Raises ValueError: as :func:seg_synapse_partners; p not strictly inside (0, 1); no end outside the background.

詳しい使い方ガイド

• emproof ファミリ ガイド

参考(サンプルデータ・文献)

• サンプルデータ カタログ(DL URL / ライセンス) — 2-D は skimage.data(BSD/public)+ 合成、3-D は実データ源(Stanford/PDS 等)の DL URL。
• 演算子の来歴・参考文献 — この op 族の元になった研究/手法の出典。
• アルゴリズムの正典(著者・年)と用途は上記ファミリ使い方ガイドに記載。

実行できる例(この op を実際に呼ぶ検証済みサンプル)

• poc_em_wiring_errors — py -3.11 examples/poc_em_wiring_errors.py

型が繋がる次の op(table を入力に取れる)

seg_synapse_partners · seg_wiring_variation · seg_synapse_nri

同カテゴリ(wiring)

seg_synapse_partners · seg_wiring_variation · seg_synapse_nri

---

Provenance: segcompare.py — EMPROOF operator registry. この per-op ノートは tools/opdocs.py md が自動生成(手編集しない)。

© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.

This site is open source. Improve this page.