Comparative Analysis of Software Solutions for Automated Recognition of Integrated Circuit Topology

The aim of this study is a comparative analysis of commercial and proprietary software tools designed for the automated recognition of integrated circuit (IC) topology elements in reverse engineering tasks. The objective is to determine the specific error profiles of each tool under consideration, namely the commercial Pix2Net software package and an author-developed Python/OpenCV code and to formulate recommendations for their improvement. The research methods include experimental recognition of IC elements fabricated using the BiCDMOS process; error analysis based on the Pareto principle with classification into three categories, namely element omission, false positive detection, and geometric error. The novelty of the work lies in comparing neural network-based and template-matching approaches to recognizing IC topology elements, revealing a trade-off between detection completeness and accuracy. The research results show that Pix2Net achieves an accuracy of 98% but generates 78.9% false positives, whereas the proprietary approach demonstrates an accuracy of 72% with omissions dominating at 60% while maintaining a low level of false detections. The conclusions indicate that each approach possesses a specific error profile. Pareto analysis allows for targeted identification of priority areas for improvement. Furthermore, the proprietary tool proves promising for educational and research purposes due to its independence from GPU hardware and commercial licenses.