{"id":"W4409262277","doi":"10.1109/wacv61041.2025.00843","title":"Unsupervised Video Highlight Detection by Learning from Audio and Visual Recurrence","year":2025,"lang":"en","type":"article","venue":"","topic":"Video Analysis and Summarization","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Saskatchewan","funders":"","keywords":"Computer science; Audio visual; Unsupervised learning; Artificial intelligence; Speech recognition; Multimedia; Computer vision","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":[0.0001561278,0.00008787456,0.0001236898,0.0001051735,0.0001921763,0.0002418216,0.0001775617,0.00005535613,0.00003579744],"category_scores_gemma":[0.0000601547,0.00007712869,0.00003307338,0.0004735883,0.00001671251,0.00039375,0.0001298004,0.00009898355,0.00001913576],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000235359,"about_ca_system_score_gemma":0.00001896017,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0003043327,"about_ca_topic_score_gemma":0.00009551936,"domain_scores_codex":[0.9991242,0.00009397191,0.0001706907,0.0003582375,0.0001334731,0.0001194596],"domain_scores_gemma":[0.9996257,0.00009524133,0.00004163761,0.000136421,0.00005824565,0.00004275989],"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":[0.00001115369,0.00006094161,0.0173932,0.00001151458,0.00007733292,0.000001911112,0.0003380789,0.0001683041,0.07613147,0.007005427,0.002800626,0.896],"study_design_scores_gemma":[0.0003018251,0.00005504787,0.01216898,0.0000214493,0.0000228778,3.494695e-7,0.00004115003,0.9333231,0.03796609,0.001341436,0.01458659,0.0001711439],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06594446,0.0003130944,0.9311268,0.0008012487,0.0001562249,0.00004553248,3.714425e-7,0.000120372,0.0014919],"genre_scores_gemma":[0.9962164,0.000108623,0.001888011,0.0002690005,0.00001937225,0.000005190935,0.0000101723,0.000002557211,0.001480729],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9331548,"threshold_uncertainty_score":0.3145217,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.004819729704766107,"score_gpt":0.2185157980074473,"score_spread":0.2136960683026812,"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."}}