wiring op• 数据种类:labels2d × labels2d × table → table
• 调用: import fullseye as fs; fs.ledger.seg_synapse_nri(a, b, synapses, spacing=None, background=0)(要直接调用实现,import segcompare; segcompare.seg_synapse_nri(a, b, synapses, spacing=None, background=0);从台账取用则 opsemproof.get("seg_synapse_nri"))
> 该算子的说明尚无译文,以下照原文给出。
Neural reconstruction integrity (Reilly et al. 2018) of candidate `b against truth a`.
NRI scores the pairs of synapse ends that a reconstruction keeps together. Over the
`2n ends (synapses as in :func:seg_synapse_partners; ends on background` in
the truth are dropped, ends on background in the candidate are singletons), a pair of
ends that lies in one truth neuron and one candidate object is a true positive; a
pair in one truth neuron but different candidate objects is a false negative (a split);
a pair in one candidate object but different truth neurons is a false positive (a
merge). `nri = 2TP / (2TP + FP + FN) with the pair precision and recall`.
Identity (checked inside, fail-closed): with `n_ij` the contingency table of the ends,
`TP = sum C(n_ij, 2), TP + FN = sum C(s_i, 2), TP + FP = sum C(t_j, 2)`, so NRI
is exactly `1 - adapted_rand_error of :func:seg_rand` taken over the ends (the
`- N form of the pair counts is 2 C(n, 2)`). Both are computed and must agree to
1e-12. Returns `nri, precision, recall, tp, fp, fn, n_ends`.
Raises `ValueError: as :func:seg_wiring_variation`; fewer than 2 ends.
• 示例数据目录(下载 URL / 许可证) —— 2-D 用 skimage.data(BSD/公有领域)加合成图,3-D 给出真实数据源(Stanford/PDS 等)的下载 URL。
• 算子来历与参考文献 —— 该算子族所依据的研究/方法出处。
• 算法的正典(作者・年份)与用途见上面的族使用指南。
• poc_em_wiring_errors — py -3.11 examples/poc_em_wiring_errors.py
table 作为输入)seg_synapse_partners · seg_wiring_variation · seg_wiring_exposure
wiring)seg_synapse_partners · seg_wiring_variation · seg_wiring_exposure
*Provenance: segcompare.py — EMPROOF 算子登记表。本条目由 tools/opdocs.py md 自动生成(请勿手工编辑)。*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.