Lens Glue Defect Detection Method Based on RevFPN

During the assembly of mobile phone lenses, dispensing defects such as glue breakage and glue overflow may occur, necessitating quality inspection of the assembled lenses. This paper proposes a RevFPN-based lens glue dispensing detection method. By utilizing a bottom-up pathway and lateral connections, the method effectively fuses high-resolution, detailrich low-level features with low-resolution but semantically strong high-level features, constructing a multi-scale feature pyramid with rich semantic information. This enhances the network's ability to detect and localize lens glue dispensing defects. Experimental results demonstrate that the proposed model achieves a detection accuracy of 94.71 %, a recall rate of 94.72 %, and a mean Intersection over Union (mIoU) of 8 3. 6 5\%, providing theoretical and technical support for lens glue dispensing defect detection.