BladeSynth: A High-Quality Rendering-Based Synthetic Dataset for Aero Engine Blade Defect Inspection

This is the authors' abstract. We don't add key points for this paper.

The integration of artificial intelligence in industry is crucial for realizing Industry 4.0; however, the lack of industrial datasets remains a significant challenge. While several generative AI methods have been proposed to create synthetic data, these approaches are often inefficient and require a large volume of training data to function effectively. In this study, we utilize a physics-based rendering procedure to generate a synthetic dataset of aeroengine blades. This dataset is then used to train a defect inspection model, thereby addressing data scarcity and enhancing defect detection accuracy in industrial applications. The dataset generation process begins with preparing Computer-Aided Design (CAD) models and material textures, then constructing a realistic inspection scene incorporating domain-randomized camera settings, lighting, and background elements. The generated data is assessed for effectiveness in both supervised and unsupervised defect detection tasks. Additionally, sim-to-real transferability is examined, demonstrating that models trained on the generated synthetic data can effectively detect and classify defects in real blade images.