{"id":"W7131093438","doi":"10.1109/iccvw69036.2025.00221","title":"Refining Naive Annotations with Limited Expert Guidance for Semantic Segmentation: A Case Study on Underwater Echograms","year":2025,"lang":"","type":"article","venue":"","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Fisheries and Oceans Canada; Canadian Water Network; ASL Environmental Sciences (Canada); University of Victoria","funders":"Alliance de recherche numérique du Canada","keywords":"Metadata; Annotation; Ground truth; Context (archaeology); Intersection (aeronautics); Artificial neural network; Segmentation; Convolutional neural network; Domain (mathematical analysis)","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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","sts"],"consensus_categories":[],"category_scores_codex":[0.0005427417,0.0005247797,0.000429787,0.0004955216,0.001573946,0.0008491701,0.0007756692,0.0001166314,0.00002981009],"category_scores_gemma":[0.0001216293,0.0004491143,0.0001247043,0.002136581,0.0001343545,0.0004457389,0.000260702,0.0004071241,0.00005717498],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002238845,"about_ca_system_score_gemma":0.0002865911,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004236736,"about_ca_topic_score_gemma":0.001093952,"domain_scores_codex":[0.9962637,0.0003379132,0.0008124395,0.001527648,0.0004432334,0.0006150483],"domain_scores_gemma":[0.9963455,0.001045517,0.000324477,0.001496826,0.0006225606,0.0001650817],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000944206,0.01604321,0.1351879,0.0005767862,0.002884595,0.002034765,0.1593094,0.135719,0.00360547,0.06159778,0.004350996,0.477746],"study_design_scores_gemma":[0.005990595,0.003587765,0.01568482,0.0005364876,0.0002419883,0.0003132873,0.02354443,0.946811,0.001074597,0.0003812108,0.0008538776,0.0009799222],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2479094,0.00005967782,0.7400578,0.007649149,0.0002271152,0.002488634,0.000009485103,0.0003246417,0.001274044],"genre_scores_gemma":[0.8338823,0.000002623474,0.1593174,0.002029107,0.00005757393,0.00227195,0.00001853078,0.00003642009,0.002384041],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.811092,"threshold_uncertainty_score":0.999796,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04203390367071803,"score_gpt":0.3603349623937063,"score_spread":0.3183010587229883,"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."}}