FLAIROP: Finalist for Best Publication at the 2022 IEEE CASE Conference
A research article from the German-Canadian project FLAIROP has made it to the final round for the "Best Conference and Application Paper Award" at the IEEE International Conference on Automation Sciences and Engineering (CASE) in Mexico City. CASE is a flagship conference of the IEEE Robotics & Automation Society. It provides a central forum for multidisciplinary research in automation and robotics.
Due to the complexity of the problem, autonomous bin picking is a major challenge for vision-guided robotic systems. In order to develop robust and effective machine learning algorithms to solve this complex task, large amounts of comprehensive and high-quality data are required. Collecting such data in the real world is too expensive and too time-consuming, and therefore not feasible from a scalability perspective. To solve this data problem, the FLAIROP team, led by the two main authors Maximilian Gilles from the Karlsruhe Institute of Technology and Yuhao Chen from the University of Waterloo, took inspiration from the concept of metaverse and presented MetaGraspNet, a large-scale photorealistic bin picking dataset constructed by physics-based metaverse synthesis. In this way, the FLAIROP team generated a dataset with more than 200 000 images created in both the metaverse and the real world. This is one of the largest datasets of its kind for the development of artificial intelligence for visually controlled robotics in production and logistics environments.
These results are summarised in the research article "MetaGraspNet: A Large-Scale Benchmark Dataset for Scene-Aware Ambidextrous Bin Picking via Physics-based Metaverse Synthesis" and can be accessed here.