{"id":"W4409263358","doi":"10.1109/wacv61041.2025.00524","title":"A Realistic Protocol for Evaluation of Weakly Supervised Object Localization","year":2025,"lang":"en","type":"article","venue":"","topic":"Advanced Image and Video Retrieval Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"École de Technologie Supérieure","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Protocol (science); Computer science; Object (grammar); Artificial intelligence; Computer vision; Medicine","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.01096878,0.002672602,0.001636657,0.00154144,0.00208119,0.001975591,0.003713025,0.003561186,0.02354086],"category_scores_gemma":[0.04076643,0.001111647,0.001073347,0.001582525,0.001880992,0.002704544,0.005265053,0.003990348,0.01925418],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002123954,"about_ca_system_score_gemma":0.00344766,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00504067,"about_ca_topic_score_gemma":0.006047734,"domain_scores_codex":[0.9880126,0.004543468,0.001245197,0.002503788,0.002740277,0.0009547907],"domain_scores_gemma":[0.9668092,0.007543071,0.001074098,0.01196579,0.01132828,0.00127947],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.007532404,0.00496572,0.01197487,0.003433856,0.0007271327,0.001504601,0.00182218,0.128553,0.1798296,0.03362468,0.2141328,0.4118991],"study_design_scores_gemma":[0.001699432,0.004285041,0.01264519,0.0006374778,0.000258639,0.001437399,0.001149655,0.4407996,0.2775089,0.06102338,0.197858,0.0006972858],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"protocol","genre_scores_codex":[0.09704608,0.0007025278,0.7785636,0.001585883,0.001791118,0.01378061,0.02682012,0.04940619,0.03030385],"genre_scores_gemma":[0.3070608,0.0004599535,0.5449864,0.002388857,0.0001944661,0.05537954,0.05263713,0.007063161,0.02982965],"genre_candidate":"protocol","genre_consensus":null,"teacher_disagreement_score":0.02354086,"threshold_uncertainty_score":0.07875198,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0539188251024046,"score_gpt":0.413897385101717,"score_spread":0.3599785599993124,"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."}}