Computer Vision: cost-effective model training with Digital Twin technology

This whitepaper is only available in English

Computer Vision

Computer vision problems can be solved using traditional vision algorithms or machine learning (ML) vision models. Traditional vision algorithms are easy to understand and have predictable results, but may struggle to perform complex tasks, while ML vision models provide high accuracy and scalability but require large datasets to be trained on, and might produce unpredictable results. This paper presents a solution to getting the training images by virtually generating synthetic training images in the game engine Unity3D.