AI-Based CT Data Pipeline

Abstract

Data generation plays an increasingly important and crucial role in training artificial intelligence (AI) models. Focusing on the field of 3D computed tomography, for example, there is often a lack of available data. To address this problem, we propose an AI-based solution to generate training data by adding artificial defects to CT data. Our method enables the generation of large amounts of realistic defect-laden parts that can be used as training data for AI applications. By automating the process and adjusting the parameters, we can generate different defect types and distributions. To evaluate the generated results, this work trains a segmentation AI and applies it to unseen real-world data. This approach closes the gap in the availability of training data and enables the industry to use AI technology effectively.