Verification picture-based multi-target data training set generation method

• Main Authors: FENG ZEHUI, ZHANG JINHONG, JIA YIZHE, QIU JIEFAN, ZHANG KUN
• Format: Patent
• Language: Chinese; English
• Published: 02.05.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention discloses a verification picture-based multi-target data training set generation method, which comprises the following steps of: identifying a preprocessed unlabeled picture by adopting a pre-trained target detection network model, segmenting a to-be-detected target region according to an identification result, and segmenting the rest part of the unlabeled picture according to a preset segmentation strategy to obtain a to-be-detected target region; obtaining an unknown segmented picture after segmentation; splicing the unknown segmented picture and the known segmented picture pushed according to the picture pushing strategy to generate a verification picture pushed to the user; and finally, after marking results of unknown segmented pictures and known segmented pictures in the verification pictures by the user are obtained, marking of unmarked pictures is realized, and the unmarked pictures are put into a training set. According to the method, the problems of low efficiency and large workload in