A robust defect detection method for syringe scale without positive samples

Abstract

With the worldwide spread of the COVID-19 pandemic, the demand for medical syringes has increased dramatically. Scale defect, one of the most common defects on syringes, has become a major barrier to boosting syringe production. Existing methods for scale defect detection suffer from large volumes of data requirements and the inability to handle diverse and uncertain defects. In this paper, we propose a robust scale defects detection method with only negative samples and favorable detection performance to solve this problem. Different from conventional methods that work in a batch-mode defects detection manner, we propose to locate the defects on syringes with a two-stage framework, which consists of two components, that is, the scale extraction network and the scale defect discriminator. Concretely, the SeNet is first built to utilize the convolutional neural network to extract the main structure of the scale. After that, the scale defect discriminator is designed to detect and label the scale defects. To evaluate the performance of our method, we conduct experiments on one real-world syringe dataset. The competitive results, that is, 99.7% on F1, prove the effectiveness of our method.

Deep Subspace K-Means Clustering for Syringe Defect Detection Without Defective Samples

“…This syringe dataset captured by industrial cameras consists of 590 normal images and 567 defective images. Following [3], these defective images are artificially generated from normal images with various kinds of defects, including "broken", "length abnormal", "excessive", "width abnormal", "missing", and "other". Some typical syringe images are shown in Fig.…”

“…The detailed procedures of the proposed method are listed as follows: Scale Extraction. We first extract the individual scales or regions of interest (ROIs) from the syringe images using the deep segmentation model in [3], only utilizing the non-defective syringe images, which can be easily collected.…”

“…Generally, clustering approaches aim to enable a significant level of similarity within individual clusters while simultaneously reducing the similarity across samples originating from different clusters. Over the course of time, there has been significant research conducted on clustering algorithms, achieving lots of successes in many domains such as data mining [1], pattern recognition [2], and defect detection [3].…”

CMC Strategies and Advanced Technologies for Vaccine Development to Boost Acceleration and Pandemic Preparedness

“…The proposed approach could represent a key asset also for the rapid response to new virus variants, allowing for a fast and optimized mRNA design. There are numerous other examples where artificial intelligence and machine learning have started bringing value to the technical development of therapeutics: from using a similar sequence modeling strategy to increase the yield of protein vaccine production in cell cultures [ 56 ] to the automated inspection of product defects based on analysis of their automatically generated images with computer-vision technology [ 57 ].…”

Deep LBP-Enriched Real-time Segmentation for Hair Defect Detection

“…Tabernik et al [16] also employed segmentationbased two-stage networks on the Kolektor surface-defect dataset and reached more advanced performance. In addition, many studies [17][18][19][20] have found that low-level visuospatial features in the network can improve the performance of semantic segmentation network. On this basis, Wang,et al [21] proposes to solve the edge blurring problem by leveraging multilevel representations during segmentation.…”