Abstract

This is an accepted article with a DOI pre-assigned that is not yet published.

Machine learning (ML), particularly within the domain of computer vision (CV), has established solutions for automated quality classification using visual data in manufacturing processes. Object detection as a CV method for quality classification provides a distinct advantage in enabling the assessment of items within the manufacturing environment, regardless of their location in images. However, substantial challenges remain regarding labeled data availability in manufacturing contexts, training examples, data imbalance, and the complexity of incorporating these methods into real-world applications. Furthermore, real-world datasets often lack adherence to FAIR principles, which limits their accessibility and interoperability, especially for small- and medium-sized enterprises (SMEs) working to integrate object detection into their manufacturing processes. In this article, we present a low-resolution 640x640 dataset based on plastic bricks for object detection, featuring two quality labels to identify minor surface defects as an example of quality classification. We analyze the dataset using a YOLOv5 model on three different dataset sizes, while accounting for class imbalance, to demonstrate the accuracy of an object detection model in a simple manufacturing use case. The mean Average Precision mAP@0.5 for correctly identifying instances in our testing dataset ranges from 0.668 to 0.774, depending on dataset size and class imbalance. While our focus is on demonstrating object detection with low-resolution images and limited data availability, the generated data and trained model also adhere to FAIR principles.Therefore, these resources are made available with proper metadata to support their reuse and further investigation into object detection tasks for similar quality classification use cases in manufacturing.