计算机工程与应用 ›› 2024, Vol. 60 ›› Issue (2): 103-112.DOI: 10.3778/j.issn.1002-8331.2208-0326

• 模式识别与人工智能 • 上一篇    下一篇

基于NVAE和OB-Mix的小样本数据增强方法

杨玮,钟名锋,杨根,侯至丞,王卫军,袁海   

  1. 1.陕西科技大学 机电工程学院,西安 710021
    2.广州先进技术研究所 机器人与智能装备中心,广州 510000
    3.中南大学 机电工程学院,长沙 410083
    4.广东技术师范大学 自动化学院,广州 510000
  • 出版日期:2024-01-15 发布日期:2024-01-15

Few Samples Data Augmentation Method Based on NVAE and OB-Mix

YANG Wei, ZHONG Mingfeng, YANG Gen, HOU Zhicheng, WANG Weijun, YUAN Hai   

  1. 1.College of Mechanical and Electrical Engineering, Shaanxi University of Science and Technology, Xi’an 710021, China
    2.Robot and Intelligent Equipment Center, Guangzhou Institute of Advanced Technology, Guangzhou 510000, China
    3.College of Mechanical and Electrical Engineering, Central South University, Changsha 410083, China
    4.College of Automation, Guangdong Polytechnic Normal University, Guangzhou 510000, China
  • Online:2024-01-15 Published:2024-01-15

摘要: 由于深度学习模型对海量标注数据的依赖性较高,导致目前许多前沿性目标检测理论难以适用于工业检测领域。为此,提出一种基于NVAE图像生成和OB-Mix数据增强的小样本数据扩充方法。具体方法是通过NVAE构建检测目标的数据分布模型,再通过采样潜变量的方式生成与真实目标图像属于同一分布的全新目标图像。在得到生成目标图像后,提出了OB-Mix数据增强策略,将生成目标图像与背景图像进行随机位置融合以构建出新的图像数据,从而提高网络的定位能力及泛化能力。方法在仅使用474张标注图像以及400张无检测目标的背景图像情况下,使YOLOv5的检测精确率达到95.86%,相比于不使用该方法的结果提高了17.60个百分点。

关键词: 数据增强, 小样本, 数据生成, 新派变分自编码器(NVAE), 表面缺陷检测, 深度学习

Abstract: Due to the high dependence of deep learning models on massive labeled data, many cutting-edge target detection theories are difficult to apply to the field of industrial detection. To this end, a small-sample data augmentation method based on NVAE for image generation and OB-Mix for data regularization is proposed. The specific method is to build a data distribution model of the detection target images through NVAE, and then generate new target images that belong to the same distribution as the real target images by sampling latent variables. After the generated target images are obtained, an OB-Mix data augmentation strategy is proposed, which mixes the generated target images with the background images at random positions to construct new images data, thereby improving the localization ability and generalization ability of the network. In the case of using only 474 labeled images and 400 background images without detection targets, the detection Precision of YOLOv5 reaches 95.86%, which is 17.60 percentage points higher than the training without this method.

Key words: data augmentation, small-sample, image generation, nouveau variational auto-encoder (NVAE), surface defect detection, deep learning