Neural Network Attention Modulation for Robust Defect Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect detection methods in industrial production face challenges in maintaining detection accuracy in suboptimal scenarios with limited training data and environmental changes such as noise, distortion, and interference.
Innovation Solution
A neural network-based defect detection system that utilizes an attention mechanism to adaptively adjust the intensity of attention maps, emphasizing defective areas and enhancing image processing through a feature extractor, attention map generator, and modulator to improve detection accuracy and generalization performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional defect detection methods are used with limited training data, then detection accuracy deteriorates, but increasing training data collection and model retraining increases time and resource consumption
Solution Approach 1:
The system performs preliminary actions by pre-processing images to enhance defective areas before detection. The image processing module pre-enhances defective areas in input images through techniques like contrast adjustment and edge enhancement, so that when the detection model processes these pre-enhanced images, it can achieve high detection accuracy without requiring extensive retraining data collection
Solution Approach 2:
The system changes parameters by dynamically adjusting image processing parameters such as contrast enhancement strength, edge detection thresholds, and attention map intensity based on the specific characteristics of each input image. This allows the detection system to adapt to different defect types and conditions without requiring model retraining, maintaining high accuracy while avoiding time-consuming data collection cycles
2Adaptability or versatility
If attention mechanism intensity is fixed in the neural network, then the system is simpler to implement, but it cannot adaptively emphasize defective areas under varying environmental conditions
Solution Approach 1:
The system implements dynamics by making the attention map intensity dynamic rather than fixed. The attention map modulator dynamically adjusts the intensity of attention maps based on the specific characteristics of each input image and the detected defect patterns. This allows the system to adaptively emphasize defective areas under varying environmental conditions such as different lighting, noise levels, and defect types, while the modular architecture keeps the implementation complexity manageable
3Measurement precision
If the system processes images without enhancing defective areas, then processing speed is faster, but detection precision decreases
Solution Approach 1:
The system applies local quality by selectively enhancing only the defective areas in images rather than processing the entire image uniformly. The image processing module identifies and enhances defective areas through local contrast adjustment and edge enhancement techniques, while leaving non-defective areas unchanged. This focused enhancement improves detection precision for defective regions without requiring extensive processing of the entire image, thereby maintaining acceptable processing speed
Data Source
Figure 1A~1B
Figure 2
Figure 3
AI summary
An apparatus including a processor configured to execute a plurality of instructions; and a memory storing the plurality of instructions, wherein execution of the plurality of instructions configures the processor to generate a defect prediction score of an input image through the use of a neural network provided reference image, the input image, and an enhanced image. The neural network may include an attention map modulator configured to adaptively adjust an intensity of an attention map generated during the use of the neural network.