P‐51: The Development Solution for Imbalanced Image Data of Cell Circuit Dents and Film Scratches of AMOLED Mass Production

AMOLED is the core technology in display industry which is made through the glass substrate process. As the AMOLED's resolution becomes higher, the glass becomes thinner, circuit of panel becomes complex and protecting film becomes important. The inspection system for securing the defects of panels and the protecting film process reliability becomes an issue in the manufacturing process. Currently, the inspection is conducted manually on random samples, thus it is impossible to inspect all the defects. for full automatic AI inspection therefor, there is a need for a cutting edge technology and high‐performance AI model which has been trained by a balanced training set. However, imbalanced data has been raised as a big issue in the mass production field. This paper proposes the novel solution for the imbalanced image data training model has been developed to automatically classify and determine the defects, thereby enhancing the detection consistency of the defects. This study applies redesigned DeepSMOTE [1] based on deep learning for an inspection system for the circuits and protection films detection process. This enhanced DeepSMOTE is capable of generating high‐quality, synthetic images can enhance minority classes and balance the training set. The effect of this AI model is to solve the imbalanced data and improve the performance of the inspection for the defects of the circuit and the film based on deep learning. Through this paper, the imbalanced image data is solved and the foundation of the solution for imbalanced image data technology applicable to mass production is established. In the test of mass production, finally the AI model got results that ‘NG Recall 98’ performance of the model. This paper proposes the enhanced the generating model in the AMOLED mass production.