Remote Sensing with limited Ground Truth

The area of remote sensing yields an abundance of multi- or even hyperspectral imagery. However, mapping these measurements to real spatial places or events is difficult and thus, ground truth is generally sparse. In this project, we examine imaging methods which require a minimum amount of ground truth.

Research Focus

We develop methods for supervised training with weak labels such as box-supervision for segmenetation or point annotations for object detection. Furthermore, we distinguish different types of errors to analyze how low-quality labels impact model performance. Finally, we also work on semi-supervised settings combining training on labeled and unlabeled data.

Publications

  • Huawei Sun, Bora Kunter Sahin, Georg Stettinger, Maximilian Bernhard, Matthias Schubert, Robert Wille: CaRaFFusion: Improving 2D Semantic Segmentation With Camera-Radar Point Cloud Fusion and Zero-Shot Image Inpainting. IEEE Robotics Autom. Lett. 10(7): 7007-7014 (2025)
  • Maximilian Bernhard, Yannic Kindermann, Roberto Amoroso, Matthias Schubert, Lorenzo Baraldi, Rita Cucchiara and Volker Tresp: What’s Outside the Intersection? Fine-grained Error Analysis for Semantic Segmentation Beyond IoU, to appear at the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV 24), 4.01–08.01 2024, Waikoloa, Hawaii
  • Maximilian Bernhard, Tanveer Hannan, Niklas Strauß, Matthias Schubert:Context Matters: Leveraging Spatiotemporal Metadata for Semi-Supervised Learning on Remote Sensing Images. ECAI 2024: 2186-2193
  • Maximilian Bernhard, Niklas Strauß, Matthias Schubert: MapFormer: Boosting Change Detection by Using Pre-change Information, in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV 23), 2-5 October 2023, Paris, France
  • Maximilian Bernhard, Matthias Schubert: Robust object detection in remote sensing imagery with noisy and sparse geo-annotations. SIGSPATIAL/GIS 2022: 25:1-25:4
  • Tanveer Hannan, Rajat Koner, Jonathan Kobold, Matthias Schubert: Box Supervised Video Segmentation Proposal Network. CoRR abs/2202.07025 (2022)
  • Maximilian Bernhard, Matthias Schubert: Robust Object Detection in Remote Sensing Imagery with Noisy and Sparse Geo-Annotations (Full Version). CoRR abs/2210.12989 (2022)
  • Maximilian Bernhard, Matthias Schubert: Correcting Imprecise Object Locations for Training Object Detectors in Remote Sensing Applications. Remote. Sens. 13(24): 4962 (2021)

What are you looking for?

More information about Google data transfer in LMU's Privacy Policy.