Fecha de publicación: 13/09/2022 Fuente: WIPO Wine The invention relates to the technical field of wine bottle surface packaging process production, in particular to a wine bottle surface defect sample generation method based on Cycle-Gan, which comprises the following steps: firstly, collecting and analyzing a surface defect sample set of a certain wine bottle; then generating a wine bottle surface defect sample by using a Cycle-Gan network on the basis of the collected and arranged defect sample set; and finally, establishing a YOLOX detection network to verify the effectiveness of the generated defect image for expansion. The invention provides a wine bottle surface defect image generation method based on a cyclic consistency generative adversarial network (Cycle-Gan) to generate a large number of defect samples so as to meet the training use requirements of a target detection neural network model. Experiments show that the performance of the detection network is effectively increased on year-on-year basis along with the increase of the scale of a data set and the quantity of samples of each defect shape tends to be uniform, and the effectiveness of the wine bottle defect generation samples based on Cycle-Gan is verified.