Maximilian Gilles, MSc

Maximilian Gilles, MSc

Robotics Research Scientist

Posts by Collection

portfolio

publications

MetaGraspNet: A Large-Scale Benchmark Dataset for Scene-Aware Ambidextrous Bin Picking via Physics-based Metaverse Synthesis

Tl;dr: Proposes the first version of MetaGraspNet dataset for robotic picking. Best Paper Award Finalist.

M. Gilles, Y. Chen, T. Robin Winter, E. Zhixuan Zeng and A. Wong, "MetaGraspNet: A Large-Scale Benchmark Dataset for Scene-Aware Ambidextrous Bin Picking via Physics-based Metaverse Synthesis," 2022 IEEE 18th International Conference on Automation Science and Engineering (CASE), Mexico-City, Mexico, August 2022, pp. 220-227, doi: 10.1109/CASE49997.2022.9926427.
Download Paper

MMRNet: Improving Reliability for Multimodal Object Detection and Segmentation for Bin Picking via Multimodal Redundancy

Tl;dr: Proposes a reliable object detection and segmentation system with Multi Modal Redundancy (MMRNet) for tackling object detection and segmentation.

Y. Chen, H. Gunraj, E. Z. Zeng, R. Meyer, M. Gilles, A. Wong, "MMRNet: Improving Reliability for Multimodal Object Detection and Segmentation for Bin Picking via Multimodal Redundancy ," IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, June 2023, pp. 68-77.
Download Paper

MetaGraspNetV2: All-in-One Dataset Enabling Fast and Reliable Robotic Bin Picking via Object Relationship Reasoning and Dexterous Grasping

Tl;dr: Contributes the second version of MetaGraspNet dataset for robotic picking by adding amodal panoptic segmentation masks and proposing an occlusion-aware fast picking method.

M. Gilles et al., "MetaGraspNetV2: All-in-One Dataset Enabling Fast and Reliable Robotic Bin Picking via Object Relationship Reasoning and Dexterous Grasping," IEEE Transactions on Automation Science and Engineering (T-ASE), 21(3), pp. 2302-2320, July 2024, doi: 10.1109/TASE.2023.3328964.
Download Paper

MetaMVUC: Active Learning for Sample-Efficient Sim-to-Real Domain Adaptation in Robotic Grasping

Tl;dr: Proposes an active learning framework using an uncertainty- and diversity-based query function for sample-efficient sim-to-real domain adaptation of picking robots.

M. Gilles, K. Furmans and R. Rayyes, "MetaMVUC: Active Learning for Sample-Efficient Sim-to-Real Domain Adaptation in Robotic Grasping," IEEE Robotics and Automation Letters, 10(4), pp.3644-3651, April 2025, doi: 10.1109/LRA.2025.3544083.
Download Paper

MPLog: A Multimodal Panoptic Dataset for Intralogistics

Tl;dr: Proposes a real-world, multi-modal (3D LiDAR and multi-view camera images) panoptic dataset for mobile robots in warehouse environments.

P. Hao, M. Gilles, D. Schüthe, S. Dierfeld, K. Furmans, "MPLog: A Multimodal Panoptic Dataset for Intralogistics ," Under Review, 2026.

research-projects

Federated Learning for Robot Picking (FLAIROP)

Funding Agency: BMWK (Germany) and NRC (Canada)
Partners: Festo SE & Co. KG (Germany), KIT (Germany), DarwinAI (Canada), University of Waterloo (Canada)
Duration: Februar 2021 - July 2023

talks

teaching