Neural Network Conversion of Circuit Diagrams into Searchable Net Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing AI systems struggle to accurately convert circuit diagrams or documents showing circuit structures into net lists due to variations in how circuit diagrams are illustrated, leading to mismatches in image searches even when the underlying specifications and structures are identical.
Innovation Solution
An AI system incorporating a neural network-based conversion mechanism that converts circuit diagrams or documents into net lists, utilizing input/output interfaces, control portions, and databases to facilitate accurate searches and retrievals of circuit structures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If image recognition processing is used to search for circuit diagrams in a database, then the search can be performed automatically, but the search accuracy deteriorates when circuit diagrams have different illustration patterns despite having the same circuit structure
Solution Approach 1:
The patent divides the circuit diagram recognition into multiple processing stages: first extracting circuit elements and their connections, then generating standardized net lists, and finally performing search based on these structured data. This segmentation allows the system to ignore visual variations and focus on structural equivalence.
Solution Approach 2:
The patent introduces a net list as an intermediary representation between the visual circuit diagram and the search database. The net list serves as a standardized intermediate format that captures circuit structure independently of illustration variations, enabling accurate matching between different diagram styles.
2Adaptability or versatility
If circuit diagrams are illustrated with different directions and positions of wiring and circuit elements, then the design flexibility is improved, but the image recognition accuracy deteriorates
Solution Approach 1:
Instead of trying to match visual patterns directly (top-down approach), the patent inverts the process by first extracting structural information and generating standardized representations (bottom-up approach). This inversion allows the system to achieve both design flexibility and recognition accuracy.
Solution Approach 2:
The patent transforms the circuit diagram from its visual representation with varying parameters (direction, position) into a standardized net list format with fixed parameter definitions. This parameter transformation preserves circuit functionality while eliminating visual variations that hinder recognition.
3Adaptability or versatility
If multiple patterns of circuit diagrams exist for the same circuit structure, then the design adaptability is improved, but the database search reliability deteriorates
Solution Approach 1:
The patent creates a universal representation format (net list) that can describe any circuit diagram pattern. This universal format serves multiple functions: it represents the circuit structure, enables standardized storage in databases, and provides consistent search criteria, thereby improving search reliability across diverse diagram styles.
Data Source
AI summary
A system that creates a net list from a circuit diagram or a document showing a circuit structure is provided. The system is an AI system including a first electronic device. The first electronic device includes an input/output interface, a control portion, and a first conversion portion. The input/output interface is electrically connected to the control portion, and the first conversion portion is electrically connected to the control portion. The input/output interface has a function of transmitting input data generated by a user's operation to the control portion, and the control portion has a function of transmitting the input data to the first conversion portion. Note that the input data is a circuit diagram illustrating a circuit structure or a document file showing the circuit structure. The first conversion portion includes a circuit where a neural network is formed, and the input data is converted to a net list with the use of the neural network of the first conversion portion.