Chip Surface Defect Detection Algorithm Based on Improved YOLOv3

• DOI: 10.1145/3638682.3638703
• Publication Date: 2024-05-22T19:32:08Z
• Authors: Biao Zhou; Xiaoyu Yu; Yang Zhang; Haisong Chen; Aili Wang

Chip package testing is a key process to eliminate bad products in the electronics industry. Aiming at problems of YOLOv3 chip detection complex environments with low accuracy and large number model parameters, an improved automatic method based on EMO, called EMO-YOLOV3, was proposed. This uses EMO replace backbone Darknet53 YOLOv3, inherits efficiency CNN short-range dependencies dynamic modeling capability Transformer learn long-distance interactions. The results show that has very good effect 7 kinds environment, such as character defect, scratch defect braid damage. Compared original model, mAP surface by about 3.4% while parameters reduced. Therefore, it considered this can be used for real-time defects.