Exploring Active Learning for Semiconductor Defect Segmentation
🤖 AI Summary
Addressing the challenges of high annotation cost, significant domain shift, and severe class imbalance—particularly for rare defect categories—in semiconductor X-ray microscopy (XRM) image defect segmentation, this paper proposes an active learning framework tailored for XRM defect segmentation. Methodologically, it integrates a semantic segmentation backbone with a class-balanced sampling mechanism. Key contributions include: (1) a contrastive learning-based cross-domain pretraining strategy to mitigate representation shift between the XRM domain and general-purpose image domains; and (2) a rarity-aware sample selection function that prioritizes both highly informative and class-sparse candidate samples. Evaluated on a high-bandwidth memory XRM dataset, the framework achieves state-of-the-art segmentation performance (a +4.2% mIoU gain), reaching full-supervision baseline accuracy using only 30% of the labeled data—substantially reducing annotation dependency.
Application Category
📝 Abstract
The development of X-Ray microscopy (XRM) technology has enabled non-destructive inspection of semiconductor structures for defect identification. Deep learning is widely used as the state-of-the-art approach to perform visual analysis tasks. However, deep learning based models require large amount of annotated data to train. This can be time-consuming and expensive to obtain especially for dense prediction tasks like semantic segmentation. In this work, we explore active learning (AL) as a potential solution to alleviate the annotation burden. We identify two unique challenges when applying AL on semiconductor XRM scans: large domain shift and severe class-imbalance. To address these challenges, we propose to perform contrastive pretraining on the unlabelled data to obtain the initialization weights for each AL cycle, and a rareness-aware acquisition function that favors the selection of samples containing rare classes. We evaluate our method on a semiconductor dataset that is compiled from XRM scans of high bandwidth memory structures composed of logic and memory dies, and demonstrate that our method achieves state-of-the-art performance.
Problem
Research questions and friction points this paper is trying to address.
Reducing annotation burden for semiconductor defect segmentation using active learning
Addressing domain shift and class-imbalance in XRM scan analysis
Improving rare class detection in semiconductor defect identification
Innovation
Methods, ideas, or system contributions that make the work stand out.
Active learning reduces annotation burden
Contrastive pretraining for AL cycle initialization
Rareness-aware acquisition for class-imbalance
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
