Automatic iterative diversity improvement for defect data generation
Open Access
- 21 June 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 41-47
Abstract
In industrial production, visual defect inspection is crucial for ensuring product quality. However, due to difficulties in sample collection during defect inspection, there is often a scarcity of diverse defective samples, leading most data generation methods to fail in generating high-diversity data. Therefore, this paper proposes the Automatic Iterative Diversity Improvement training strategy to improve the diversity of generated data with only a few training samples, which can improve the performance and generalization of downstream tasks. During the training process, given the characteristics of visual defects, we design the Defect Diversity-based Picking and Evaluation to pick up diverse samples and assess the diversity regarding defect location and shape. Experiments show that the method can generate high-diversity defect samples with high quality for industrial defect scenarios even in zero-shot learning tasks. The effectiveness of the generated data is also verified by the down-streaming defect detection task.Keywords
This publication has 12 references indexed in Scilit:
- YOLOV5-based traffic sign detection algorithm2023
- A Fabric Defect Detection Method Based on Deep LearningIEEE Access, 2022
- Region- and Strength-Controllable GAN for Defect Generation and Segmentation in Industrial ImagesIEEE Transactions on Industrial Informatics, 2021
- ILVR: Conditioning Method for Denoising Diffusion Probabilistic Models2021
- Few-Shot Defect Segmentation Leveraging Abundant Defect-Free Training Samples Through Normal Background Regularization And Crop-And-Paste Operation2021
- A survey of deep learning techniques for weed detection from imagesComputers and Electronics in Agriculture, 2021
- An End-to-End Steel Surface Defect Detection Approach via Fusing Multiple Hierarchical FeaturesIEEE Transactions on Instrumentation and Measurement, 2019
- The Unreasonable Effectiveness of Deep Features as a Perceptual Metric2018
- A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defectsApplied Surface Science, 2013
- Image quality assessment: from error visibility to structural similarityIEEE Transactions on Image Processing, 2004