IC Pattern Defect Detection With Attention-Guided Image Context
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect detection methods for objects with integrated circuit patterns, such as photolithography masks and wafers, face challenges in accuracy and efficiency due to limitations in considering global image context and the fixed weights of convolutional neural networks.
Innovation Solution
A computer-implemented method using a machine learning model with an attention mechanism for defect detection in imaging datasets of objects with integrated circuit patterns, which enhances defect highlighting and improves accuracy by considering the structural context of circuit patterns.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If convolutional neural networks with fixed weights are used for defect detection, then the detection process is computationally efficient, but the accuracy is limited due to inability to consider global image context
Solution Approach 1:
The patent segments the defect detection task into two distinct stages: a coarse detection stage using a lightweight CNN for initial defect localization, and a fine detection stage using a Transformer model with attention mechanisms for precise defect characterization. This segmentation allows the system to benefit from both the computational efficiency of CNNs and the global context understanding of Transformers, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The patent introduces dynamic adaptability by making the second stage model configurable based on the detection needs. The system can dynamically adjust whether to apply the computationally intensive Transformer model only to regions containing detected defects or to the entire image, thereby balancing accuracy requirements with computational resources and model complexity.
2Reliability
If traditional defect detection methods are used, then the computation time is reduced, but the accuracy and reliability of defect detection deteriorates
Solution Approach 1:
The patent applies preliminary action by using the first stage CNN model to pre-process the image and identify potential defect regions before applying the more reliable but computationally intensive Transformer model. This preliminary filtering ensures that the high-accuracy model only processes relevant regions, maintaining reliability while significantly reducing overall computation time compared to applying the Transformer model to the entire image.
3Adaptability or versatility
If advanced machine learning models with attention mechanism are applied, then the adaptability to different applications and imaging datasets is improved, but the memory requirements and user effort increase
Solution Approach 1:
The patent implements dynamic resource allocation where the system adaptively determines whether to apply the full two-stage process or simplified versions based on dataset characteristics and computational constraints. This allows high adaptability across different applications while dynamically adjusting memory consumption and user effort requirements to match the specific task demands.
Data Source
AI summary
The invention relates to a computer implemented method for defect detection comprising: obtaining an imaging dataset and a reference dataset of an object comprising integrated circuit patterns; and detecting defects in the imaging dataset using the imaging dataset and the reference dataset, wherein a machine learning model for defect highlighting is applied to the imaging dataset as input and generates a highlighted defect dataset as output, and wherein the machine learning model for defect highlighting comprises at least one attention mechanism. The invention also relates to computer programs, computer-readable media and corresponding systems.