{"id":"W4410827183","doi":"10.1007/s10845-025-02623-3","title":"Focus to generalize (F2G): physics-guided attention for sample efficient and generalizable deep learning defect detection","year":2025,"lang":"en","type":"article","venue":"Journal of Intelligent Manufacturing","topic":"Industrial Vision Systems and Defect Detection","field":"Engineering","cited_by":1,"is_retracted":false,"has_abstract":false,"ca_institutions":"National Research Council Canada; University of Toronto","funders":"","keywords":"Focus (optics); Sample (material); Artificial intelligence; Deep learning; Computer science; Cognitive science; Machine learning; Engineering; Psychology; Physics; Optics","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0005673099,0.0001876414,0.0003106552,0.0003568599,0.0001956609,0.0001270738,0.00008499395,0.0001123928,0.000009779421],"category_scores_gemma":[0.0001347863,0.0001705958,0.0002559232,0.0001680687,0.000008304428,0.0001004022,0.00003149498,0.0002514559,0.000004747983],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000263801,"about_ca_system_score_gemma":0.00001184024,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006406887,"about_ca_topic_score_gemma":0.0000174954,"domain_scores_codex":[0.9987224,0.00005590246,0.0006115239,0.0001645619,0.0001896888,0.0002558952],"domain_scores_gemma":[0.999343,0.0001492731,0.0001652812,0.0001068874,0.0001375537,0.00009802477],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.00006402294,0.00001238216,0.00008109031,0.00008780569,0.0001146037,0.00000151442,0.0001230853,0.7769904,0.03863199,0.00003988565,0.0001266767,0.1837265],"study_design_scores_gemma":[0.0005489778,0.0002022768,0.0002990689,0.0001898037,0.00008540388,0.00003661761,0.0001475219,0.1136093,0.8689573,0.0003082833,0.01541795,0.0001974807],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4584737,0.0002706908,0.5398709,0.00001207295,0.001045697,0.0002097603,0.000001104995,0.00003955627,0.00007658382],"genre_scores_gemma":[0.9968944,0.00007137784,0.002290845,0.00002125508,0.0005334786,0.00001874915,0.000001333264,0.00002904403,0.0001394906],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8303254,"threshold_uncertainty_score":0.6956694,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01811174095409849,"score_gpt":0.2584816367091504,"score_spread":0.2403698957550519,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}