MASTERLY_niMble Artificial intelligence driven robotic SoluTions for Efficient and self-deteRmined handLing and assemblY operations (G.A.101091800)
Published March 22, 2025 | Version v1
Conference proceeding Open

Leveraging Generative AI for Synthetic Data Generation: Improving 6-DOF Pose Estimation in Assembly Systems

Authors/Creators

  • 1. Laboratory for Manufacturing Systems and Automation (LMS)

Description

In this paper, a synthetic dataset enhanced by GAN-generated textures is presented to improve pose estimation. The distinguish feature compared with novel 6-DOF pose estimation models, is the utilization of GAN generated content that augmented key characteristics of the approach, mimicing real industrial parts, and environments. The proposed method was tested and evaluated in a real industrial use case derived from MASTERLY KLEEMANN pilot, consisting of a set of 3 different components: a term block, a circuit breaker and a relay switch.

Files

978-3-031-86489-6.pdf

Files (39.2 MB)

Name Size Download all
md5:e890d34c179a1b6116be63de04a78c7c
39.2 MB Preview Download

Additional details

Identifiers

ISBN
978-3-031-86489-6

Funding

European Commission
MASTERLY - Nimble Artificial Intelligence driven robotic solutions for efficient and self-determined handling and assembly operations 101091800

Software

Repository URL
https://link.springer.com/chapter/10.1007/978-3-031-86489-6_8#Bib1

References

  • Abufadda, M., Mansour, K.: A survey of synthetic data generation for machine learning. In: 2021 22nd International Arab Conference on Information Technology (ACIT), pp. 1–7. IEEE (2021)
  • Cao, H., Dirnberger, L., Bernardini, D., Piazza, C., Caccamo, M.: 6IMPOSE: bridging the reality gap in 6d pose estimation for robotic grasping (Mar 2023). http://arxiv.org/abs/2208.14288, arXiv:2208.14288
  • Chen, W., Jia, X., Chang, H.J., Duan, J., Leonardis, A.: G2L-Net: Global to Local Network for Real-time 6D Pose Estimation with Embedding Vector Features (2020). https://arxiv.org/abs/2003.11089
  • Chryssolouris, G.: Manufacturing systems: theory and practice. Mechanical engineering series, Springer, New York, 2nd ed edn. (2006), oCLC: ocm61253973
  • Denninger, M., Winkelbauer, D., Sundermeyer, M., Strobl, K.H., Humt, M., Triebel, R.: BlenderProc2: A procedural pipeline for photorealisticrendering. J. Open Source Softw. 8(82), 4901 (2023). https://doi.org/10.21105/joss.04901
  • Labbé, Y., et al.: MegaPose: 6D Pose Estimation of Novel Objects via Render & Compare (Dec 2022). http://arxiv.org/abs/2212.06870, arXiv:2212.06870 [cs]
  • Lin, J., Liu, L., Lu, D., Jia, K.: SAM-6D: Segment Anything Model Meets Zero-Shot 6D Object Pose Estimation (Mar 2024). http://arxiv.org/abs/2311.15707, arXiv:2311.15707 [cs]
  • Makris, S.: Cooperating Robots for Flexible Manufacturing. Springer International Publishing, Cham (2021)
  • Peebles, W., Xie, S.: Scalable Diffusion Models with Transformers (Mar 2023). arXiv:2212.09748 [cs]
  • Redmon, J., Divvala, S., Girshick, R., Farhadi, A.: You Only Look Once: Unified, Real-Time Object Detection (May 2016). arXiv:1506.02640 [cs]
  • Rojtberg, P., Pollabauer, T., Kuijper, A.: Style-transfer GANs for bridging the domain gap in synthetic pose estimator training. In: 2020 IEEE International Conference on AIVR, pp. 188–195. IEEE, Utrecht, Netherlands (Dec 2020). https://doi.org/10.1109/AIVR50618.2020.00039
  • Tremblay, J., To, T., Sundaralingam, B., Xiang, Y., Fox, D., Birchfield, S.: Deep Object Pose Estimation for Semantic Robotic Grasping of Household Objects (Sep 2018). arXiv:1809.10790 [cs]
  • Wen, B., Yang, W., Kautz, J., Birchfield, S.: FoundationPose: Unified 6D Pose Estimation and Tracking of Novel Objects (2023). https://doi.org/10.48550/ARXIV.2312.08344, https://arxiv.org/abs/2312.08344
  • Xiang, Y., Schmidt, T., Narayanan, V., Fox, D.: PoseCNN: a convolutional neural network for 6D object pose estimation in cluttered scenes (May 2018). arXiv:1711.00199 [cs]