Abstract: This paper presents a systematic method to create universally applicable synthetic training image datasets for computer vision-based construction object detection. The synthetic images created by inserting a virtual object ...