read_pod op• Data kinds: table → table
• Call: import fullseye as fs; fs.ledger.symbol_read_pod(trials_table, *, link='logit', level=0.95) (to call the implementation directly, import symbolpod; symbolpod.symbol_read_pod(trials_table, *, link='logit', level=0.95); from the registry, opssymbol.get("symbol_read_pod"))
• GPU: none (CPU reference implementation only)
From a table of read trials, return the POD (hit = the true string was read) and the misread upper bound (table).
> The detailed description below is the original text — the summary and the headings are translated.
Fit the hit/miss POD (pod.pod_hitmiss_fit) to a table from symbol_read_trials. When the factor is worse when larger (defocus, noise …) the fit runs on x = −level so that POD rises with x, and the answers are mapped back: level90 = the degradation read with 90 % probability, level90_95 = its conservative bound at level confidence (a *smaller* degradation), level50 likewise. When the factor is better when larger (module size, contrast, full well) level90_95 is the *larger* value.
method = "hitmiss_glm" then. ★When the outcome is decided by the level (complete separation — every capture read up to one level and none beyond), the slope is not estimable; the op then returns method = "all_success_demonstration": the most severe level at which every trial read (separation_bracket = that level and the first one with a failure), and level90_95 = that level only if it holds at least pod.trials_for_all_success (0.90, level) trials (29 at 95 %), else None. level50 / level90 are None in that case.
★monotone is False when a more degraded level reads more often than a milder one by more than 3 standard errors (nonmonotone_z): the POD models assume monotone degradation, so the fitted numbers are then outside their premise (seen with a reader that cannot read sharp dot-peen marks but reads them once blur merges the dots). The result is still returned, flagged.
Also returns per_level (level, n, hits, misreads, noreads, read_rate), misread_total / decoded_total and misread_rate_upper (one-sided Clopper-Pearson upper bound of misreads per decode at level; with zero misreads in n decodes it is 1 − (1 − level)^(1/n)), fit (the POD table) and factor.
空の入力(0 行)は ValueError。拒否(ValueError): 列の欠け、全部読めた/全部読めない、level が (0, 1) の外。完全分離は拒否せず上の「全数成功の実証」で返す(傾きの無い POD を作らない)。
• 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.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• poc_symbol_lot_ocr — py -3.11 examples/poc_symbol_lot_ocr.py
• poc_symbol_package_1d — py -3.11 examples/poc_symbol_package_1d.py
• poc_symbol_part_dpm — py -3.11 examples/poc_symbol_part_dpm.py
table as input)render_symbol · symbol_read_trials · ocr_classify · ocr_read_line · ocr_read_trials
read_pod)symbol_read · symbol_read_trials
*Provenance: symbolpod.py — SYMBOL 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.