Chip Defect Detection Using ACmix and Attention Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current chip defect detection methods, particularly using target detection networks, suffer from low accuracy due to limited feature extraction capabilities and low attention to small-size targets.
Innovation Solution
A method utilizing a trained target detection network that incorporates a long-range attention network, squeeze-and-excitation network, and an ACmix module to enhance feature extraction and attention to small-size targets, improving defect detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a conventional target detection network is used for chip defect detection, then the detection process is simple, but the accuracy is low due to limited feature extraction capability and low attention to small-size targets
Solution Approach 1:
The detection network is segmented into multiple specialized modules: a backbone network for feature extraction, a long-range attention network for global context modeling, a squeeze-and-excitation network for channel-wise feature recalibration, and an ACmix module for small target enhancement. Each module focuses on specific aspects of feature processing to collectively improve detection accuracy.
Solution Approach 2:
The network employs nested module structures where the squeeze-and-excitation network is embedded within the feature extraction pipeline, the long-range attention network operates on top of backbone features, and the ACmix module integrates multiple feature pyramids. These nested structures allow progressive refinement of features at different levels.
2Measurement precision
If the target detection network focuses on general features, then the overall detection coverage is maintained, but the attention to small-size defects is insufficient
Solution Approach 1:
The ACmix module implements local quality enhancement by processing different feature pyramids with specialized operations tailored to small target characteristics. The module applies localized attention mechanisms that prioritize small defect regions while preserving overall feature integrity through multi-scale feature fusion.
Solution Approach 2:
The network addresses small target detection by adding the dimension of multi-scale feature processing. The ACmix module operates in feature space by combining features from multiple scales and resolutions, effectively creating an additional dimensional perspective that highlights small defects without losing global context.
Data Source
AI summary
A method for defect detection of a chip, an electronic device, and a storage medium. The method includes: obtaining a surface image and a package image of the chip; and inputting the surface image and the package image into a trained target detection network for performing the defect detection, thereby obtaining a surface defect and a package defect of the chip, respectively. The target detection network includes a long-range attention network and a squeeze-and-excitation network configured for determining feature weights of multiple feature channels, and an ACmix module configured for locating and recognizing a small-size target in an image. The method for defect detection of the chip is applied to enable a capability of extracting important features in the chip to be much stronger, enable the attention to the defect in the chip to be much higher, and improve an accuracy of defect detection of the chip.