Neural network and image processing-based automatic netlist generation for complex analog circuits

In the field of electronic design automation (EDA), the rapid development of artificial intelligence has automated many manual and time-consuming tasks in circuit analysis and simulation. However, generating netlists from complex analog circuit diagrams remains challenging due to diverse component types, intricate connectivity relationships, and significant variations in diagram styles. In this regard, this paper proposes an intelligent solution based on artificial intelligence and image processing to realize the full-process automation of analog circuits from circuit diagrams to netlists. In this paper, the YOLOv11 model is employed to accurately detect and locate the components in the circuit diagram, and combined with the connectivity domain detection and multiangle template matching technology, the pin functions of the components and the circuit connection relationship are deduced, so as to realize the automation of the conversion from the circuit diagram to the netlist. The experimental results show that the method achieves a component recognition accuracy of 97.8% after training on a dataset of 1994 circuit diagrams. Furthermore, it generates fully correct netlists for 94.7% of 60 circuit diagrams annotated with reference netlists, with an average graph edit distance (GED) of 13.3 for incorrect netlists.