seg_rand — EMPROOF `score` op

• 資料種類:labels2d × labels2d → table

• 呼叫: import fullseye as fs; fs.ledger.seg_rand(a, b, ignore_label=None)(要直接呼叫實作,import segcompare; segcompare.seg_rand(a, b, ignore_label=None);從台帳取用則 opsemproof.get("seg_rand"))

用法

> 該運算子的說明尚無譯文,以下照原文給出。

Rand index, adjusted Rand index and the adapted Rand error (CREMI / SNEMI).

With `n_ij the contingency table, a_i / b_j its row and column sums and N`

the pixel count:

• `rand_index` = fraction of pixel pairs on which the two agree (same/different);

• `adjusted_rand_index` = Hubert & Arabie's chance-corrected version (1 = identical,

about 0 = independent);

• `adapted_rand_error = 1 - F over **pairs of distinct pixels**: pair precision` =

(sum n_ij^2 - N) / (sum b_j^2 - N) — of the pairs the candidate puts together, the

fraction the truth also puts together — and `recall` = (sum n_ij^2 - N) /

(sum a_i^2 - N) (Arganda-Carreras et al. 2015). The `- N` removes each pixel's pair

with itself. scikit-image's `adapted_rand_error` returns the same error, but its

`precision` is divided by the truth pairs (its code sums the rows of the

truth-by-test table), i.e. it is this `recall`; its docstring says otherwise.

Raises `ValueError: as :func:seg_contingency`; fewer than 2 counted pixels.

詳細使用指南

• emproof 族使用指南

參考(範例資料・文獻)

• 範例資料目錄(下載 URL / 授權) —— 2-D 用 skimage.data(BSD/公有領域)加合成圖,3-D 給出真實資料源(Stanford/PDS 等)的下載 URL。

• 運算子來歷與參考文獻 —— 該運算子族所依據的研究/方法出處。

• 演算法的正典(作者・年份)與用途見上面的族使用指南。

可執行的範例(實際呼叫該運算子並已驗證的樣例)

• poc_em_split_merge_score — py -3.11 examples/poc_em_split_merge_score.py

型別可銜接的下一個運算子(可接受 `table` 作為輸入)

seg_synapse_partners · seg_wiring_variation · seg_synapse_nri · seg_wiring_exposure

同類別(`score`)

seg_contingency · seg_variation_of_information

*Provenance: segcompare.py — EMPROOF 運算子登記表。本條目由 tools/opdocs.py md 自動產生(請勿手動編輯)。*

seg_rand — EMPROOF score op

用法