Research on PCB Defect Detection Based on Improved YOLOv8

Reliable PCB inspection must identify defects that occupy only a few pixels while separating categories with highly similar visual patterns. To address the resulting loss of detail and unstable localization, an improved YOLOv8 detector is developed with a high-resolution small-target branch, a reduced large-target branch, BiFormer contextual attention, and WIoU v3 regression. The design is evaluated on PKU-PCB and additionally tested on Deep PCB. On PKU-PCB, it obtains 97.11% mAP@0.5 and 94.81% Recall, improving mAP@0.5 by 4.40 percentage points over YOLOv8s. The detector contains 8.38 M parameters, 25.14% fewer than YOLOv8s (11.2 M), and reaches 68 FPS under the reported protocol. The ablation results show that the high-resolution branch provides the main recall gain, whereas BiFormer and WIoU v3 further improve contextual discrimination and box regression. These results indicate that targeted structural reduction can improve both detection accuracy and computational efficiency for PCB inspection.