{"id":"W4320518953","doi":"10.1016/j.autcon.2023.104771","title":"Hybrid DNN training using both synthetic and real construction images to overcome training data shortage","year":2023,"lang":"en","type":"article","venue":"Automation in Construction","topic":"3D Surveying and Cultural Heritage","field":"Earth and Planetary Sciences","cited_by":48,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Toronto","funders":"","keywords":"Economic shortage; Artificial intelligence; Training (meteorology); Field (mathematics); Computer science; Automation; Artificial neural network; Synthetic data; Training set; Machine learning; Engineering","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":[0.0007799463,0.0001397262,0.0001944833,0.0002454969,0.000251814,0.0001440076,0.0001330452,0.00005956161,0.0001784621],"category_scores_gemma":[0.0001458264,0.0001380084,0.00002046454,0.0004550928,0.0001313612,0.0008182562,0.00002557139,0.0001221958,0.00003751259],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001728848,"about_ca_system_score_gemma":0.00006336773,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001484915,"about_ca_topic_score_gemma":0.000693986,"domain_scores_codex":[0.9986084,0.0001485738,0.0003316852,0.0004095236,0.0002181334,0.0002836814],"domain_scores_gemma":[0.9993821,0.0001786515,0.00009577093,0.0002243457,0.00002385138,0.0000952344],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.00001600316,0.000002503743,0.266773,0.00003057883,0.00001298362,0.00001829525,0.002307779,0.002364986,0.001038901,0.000112062,0.00006845767,0.7272545],"study_design_scores_gemma":[0.0002532077,0.00003420913,0.7584251,0.0001231096,0.00001664565,0.0004454897,0.008427863,0.2310173,0.0001026009,0.0007363206,0.0001563652,0.0002617747],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9966378,0.00003750448,0.0008316808,0.0001069349,0.0007130988,0.0001687553,0.0002132574,0.0002773465,0.0010136],"genre_scores_gemma":[0.9884881,0.00005075366,0.01095417,0.00002604296,0.00009522311,0.000001742633,0.000354467,0.000006069255,0.0000233881],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7269927,"threshold_uncertainty_score":0.5627818,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09673258083927715,"score_gpt":0.2900833869106985,"score_spread":0.1933508060714214,"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."}}