While deep neural networks have achieved remarkable performance, data augmentation has emerged as a crucial strategy to mitigate overfitting and enhance network performance. These techniques hold particular significance in industrial manufacturing contexts. Recently, image mixing-based methods have been introduced, exhibiting improved performance on public benchmark datasets. However, their application to industrial tasks remains challenging. The manufacturing environment generates massive amounts of unlabeled data on a daily basis, with only a few instances of abnormal data occurrences. This leads to severe data imbalance. Thus, creating well-balanced datasets is not straightforward due to the high costs associated with labeling. Nonetheless, this is a crucial step for enhancing productivity. For this reason, we introduce ContextMix, a method tailored for industrial applications and benchmark datasets. ContextMix generates novel data by resizing entire images and integrating them into other images within the batch. This approach enables our method to learn discriminative features based on varying sizes from resized images and train informative secondary features for object recognition using occluded images. With the minimal additional computation cost of image resizing, ContextMix enhances performance compared to existing augmentation techniques. We evaluate its effectiveness across classification, detection, and segmentation tasks using various network architectures on public benchmark datasets. Our proposed method demonstrates improved results across a range of robustness tasks. Its efficacy in real industrial environments is particularly noteworthy, as demonstrated using the passive component dataset.

Introduction

Modern industrial manufacturing environments have been utilizing various machine-learning techniques to increase the efficiency of production procedures. In particular, the techniques using image processing, such as those proposed by Guerra et al. (1997), Lyu and Chen (2009), and Rejc et al. (2011) apply to detect defects of goods or parts, reduce human error, and improve the quality of products. Deep neural networks (DNNs) have shown incomparable performances on various computer vision tasks, which are image classification (He et al., 2016a, Han et al., 2017, Xie et al., 2017), object detection (Ren et al., 2015, Liu et al., 2016), and semantic segmentation (Long et al., 2015, Chen et al., 2018, Zhang et al., 2020). Then, various approaches have been proposed to improve the model performance by designing more complex but efficient network architectures or gathering more large-scale datasets such as ImageNet (Krizhevsky et al., 2012), YFCC-100M (Thomee et al., 2016), and IG-1B-Targeted (Yalniz et al., 2019). In this regard, various techniques (Zhang et al., 2019, Yang et al., 2021) have adopted the DNNs to visual inspection systems in the manufacturing process to improve productivity.

Read the full article at ScienceDirect.com.