In today's dynamic industrial landscape, where products are increasingly diverse and complex, the need for adaptive robotic sensing is constantly rising. Leveraging deep learning for image or point cloud segmentation enables complex and adaptive assembly and production.

However, the challenge lies in acquiring vast amounts of annotated data to train these AI algorithms [2,3]. Traditionally, this has been a resource-intensive endeavour. But here's where the magic begins: the concept of 'sim2real'—bridging the gap between simulation and reality [3]. By generating synthetic data, we can fast-track the training process without relying solely on real-world data.

And that's precisely the focus of our project: pioneering an automated data generation system tailored specifically for the intricate realm of industrial assembly. Join us as we push the boundaries of innovation and redefine the future of robotic perception in industrial settings.