Artificial Dataset Generation for Automated Aircraft Visual Inspection Artificial Dataset Generation for Automated Aircraft Visual Inspection

Aircraft visual inspection is both essential to the maintenance of an aircraft, and expensive and time-consuming to perform. Augmenting trained maintenance professionals with automated UAVs to collect and analyze images for aircraft inspection is an active research topic and a potential application of convolutional neural networks (CNNs). Training datasets for niche research topics such as aircraft visual inspection are small and challenging to produce, and the manual labeling process of these datasets produces subjective annotations. Self-driving car researchers have experimented with generating artificial datasets with modern computer graphics that can train for real-world driving scenarios. Our research borrows this idea and proposes a work-in-progress artificial data generation pipeline to create 3D rendered automatically annotated images for training CNNs for automated visual aircraft inspection.

Automated summary

Aircraft visual inspection is crucial but time-consuming and expensive. Researchers are exploring the use of UAVs and CNNs for automated inspection, but face challenges in generating training datasets due to their niche nature. Innovative approaches, such as leveraging artificial data generation techniques from self-driving car research, are being proposed to address this issue.