Synthetic Data Generation for Surface Defect Detection

Abstract

Ensuring continued quality is challenging, especially when customer satisfaction is the provided service. It seems to become easier with new technologies like Artificial Intelligence. However, field data are necessary to design an intelligent assistant but are not always available. Synthetic data are used mainly to replace real data. Made with a Generative Adversarial Network or a rendering engine, they aim to be as efficient as real ones in training a Neural Network. When synthetic data generation meets the challenge of object detection, its capacity to deal with the defect detection challenge is unknown. Here we demonstrate how to generate these synthetic data to detect defects. Through iterations, we apply different methods from literature to generate synthetic data for object detection, from how to extract a defect from the few data we have to how to organize the scene before data synthesis. Our study suggests that defect detection may be performed by training an object detector neural network with synthetic data and gives a protocol to do so even if at this point, no field experiments have been conducted to verify our detector performances under real conditions. This experiment is the starting point for developing a mobile and automatic defect detector that might be adapted to ensure new product quality.