arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.05897cs.CV
收藏

基于领域差异的自动光学检测模型基迁移学习

Model-based Transfer Learning for Automatic Optical Inspection based on domain discrepancy

Tsinghua-Berkeley Shenzhen Institute(清华-伯克利深圳学院);Shenzhen Han Industrial Technologies Co., Ltd.(深圳汉工业技术有限公司);Guilin Han Smart Instrument Co., Ltd.(桂林汉智能仪器有限公司);Guangxi Guihua Intelligent Manufacturing Co., Ltd.(广西桂华智能制造有限公司)

Erik Isai Valle Salgado, Haoxin Yan, Yue Hong, Peiyuan Zhu, Shidong Zhu, Chengwei Liao, Yanxiang Wen, Xiu Li, Xiang Qian, Xiaohao Wang, Xinghui Li

摘要:

迁移学习是AOI应用的一种很有前途的方法,因为它可以显着缩短样本收集时间,提高当今智能制造的效率。然而,相关研究通过应用TL来增强网络模型,而没有考虑数据集之间的域相似性,源数据集的数据长尾性,并且主要使用线性变换来缓解样本不足。本研究应用基于模型的TL通过领域相似性来提高整体性能,并在目标和源领域进行数据增强,以丰富数据质量并减少不平衡。给定一组来自相似工业过程的源数据集,我们通过域差异分数和每个样本的数量来定义哪一组与目标最相关。然后,我们将选择的预训练骨干权重传递给训练并微调目标网络。我们的研究表明,与使用基准数据集的TL相比,F1评分和PR曲线增加了20%。

英文摘要:

Transfer learning is a promising method for AOI applications since it can significantly shorten sample collection time and improve efficiency in today's smart manufacturing. However, related research enhanced the network models by applying TL without considering the domain similarity among datasets, the data long-tailedness of a source dataset, and mainly used linear transformations to mitigate the lack of samples. This research applies model-based TL via domain similarity to improve the overall performance and data augmentation in both target and source domains to enrich the data quality and reduce the imbalance. Given a group of source datasets from similar industrial processes, we define which group is the most related to the target through the domain discrepancy score and the number of samples each has. Then, we transfer the chosen pre-trained backbone weights to train and fine-tune the target network. Our research suggests increases in the F1 score and the PR curve up to 20% compared with TL using benchmark datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑