DeepICLogo: A Novel Benchmark Dataset for Deep Learning-Based IC Logo Detection

The growing number of hardware problems caused by undetected faults during PCB fabrication is concerning. Manufacturers need more potent quality control processes to minimize these challenges. Logo detection and verification play a critical role in ensuring quality control and detecting counterfeits in the field of PCB assurance. In this manuscript, the importance of IC Logo Detection in Automated Visual Inspection (AVI) and PCB assurance along with its existing challenges and current solutions has been discussed. In addition to it, we present and describe a novel IC logo dataset and its contributions to Deep learning-based Logo Detection. We implemented the transfer learning technique using the state-of-the-art You Only Look Once (YOLO) model, more specifically, YOLOv5 model, and discussed the results and associated insights. We have also suggested the possible improvements required for future work.