anomaly_eval op• Data kinds: video × video → table
• Call: import fullseye as fs; fs.ledger.anomaly_pixel_f1_max(scores, masks) (to call the implementation directly, import anomeval; anomeval.anomaly_pixel_f1_max(scores, masks); from the registry, opsspc.get("anomaly_pixel_f1_max"))
Best pixel-level F1 over all thresholds of anomaly score maps against defect masks (table).
scores and masks are (n, H, W) (or lists / single 2-D maps); a pixel is predicted defective when its score is >= the threshold. The F1 = 2TP / (2TP + FP + FN) is evaluated exactly at every distinct score (sorted cumulative sums, no threshold grid); the highest threshold wins a tie. Returns f1, threshold, precision and recall at the maximum, n_pos and n_neg (pixels). It counts pixels, like the pixel AUROC, not regions (see anomaly_pro): a small defect weighs little however well it is found.
Raises ValueError on empty / mismatched / non-finite input, non-binary masks, or when the masks contain no defect pixel or no good pixel.
• 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_patch_anomaly — py -3.11 examples/poc_patch_anomaly.py
table as input)msa_anova_table · msa_gauge_rr · msa_bias_linearity · msa_attribute_agreement · gum_standard_uncertainty · gum_propagate · gum_expanded · gum_monte_carlo
anomaly_eval)anomaly_image_auroc · anomaly_pixel_auroc · anomaly_pro · anomaly_region_overlap · anomaly_image_f1_max · anomaly_operating_point
*Provenance: anomeval.py — SPC 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.