Automated Netlist Generation from Offline Hand-Drawn Circuit Diagrams
Waqas Uzair, Douglas Chai, Alexander Rassau · 2023
This paper introduces a novel method for the precise and efficient interpretation of hand-drawn circuit diagrams. This method addresses the challenge of understanding electrical drawings by not only classifying and localizing electronic components, as previous works have done, but also recognizing textual labels and their associations with electronic components. The method addresses the complexities of recognizing decimal points, grouping textual labels, and associating the grouped textual labels with the corresponding electronic components. These complexities arise due to the relative locations of individual textual labels, the orientation of textual labels, and the location of electronic components in the circuit. To evaluate the proposed method, we tested it on a set of 100 randomly selected images. The method correctly identified the objects in 87% of the images and accurately generated the Netlist for 82% of the images. This resulted in an accuracy rate of 94.24% for the object identification algorithm and an overall accuracy rate of 82% for the method.