Clustering algorithms that rely on local density estimates frequently suffer from bridges of density-connected points between clusters that have no real underlying relationship. Such bridges may arise from touching, overlapping, and noise-induced clusters, leading to falsely connected clusters. Although approaches have been developed to tackle this problem, they introduce new challenges, including high computational complexity, the need for hyperparameters, and unstable performance for a wider range of clustering problems. We present a clustering algorithm that incorporates the novel topological feature lifetime by iteratively contracting a fixed-radius neighbor graph. The lifetime aggregates information from outside the local neighborhood of points by capturing the influence of the global contraction sequence induced by the initial graph topology. The difference in lifetimes between a node and its neighbors reveals topological centers and inevitably associated connected components. Accordingly, we determine a cluster skeleton and then assign the remaining nodes to the dominant cluster in their neighborhood. The algorithm requires only the radius within a point to be considered a neighbor as input, is free of additional hyperparameters, and has a time complexity no higher than that of related algorithms while computing a global feature. Experiments on diverse benchmark datasets demonstrate that our algorithm is robust to typical clustering problems. Our implementation achieves highly competitive performance relative to the state-of-the-art while requiring less runtime and memory than related approaches.

Bavaria Buildings is a large, analysis-ready dataset providing openly available co-registered 40cm aerial imagery of Upper Bavaria paired with building footprint information. The Bavaria Buildings dataset (BBD) contains 18205 orthophotos of 2500 × 2500 pixels, where each pixel covers 40cm × 40cm in space (Digitales Orthophoto 40cm - DOP40). The dataset has been pre-processed and co-registered and also provides a set of 5.5 million image tiles of 250 × 250 pixels ready for deep learning and image analysis tasks. For each image tile, we provide two segmentation masks; one based on the official building footprints (Hausumringe) data as published by the Free State of Bavaria and one based on a historic OpenStreetMap (OSM) extract dating to 2021. The dataset is ready for essential analysis tasks, such as detection, segmentation, instance extraction, footprint geometry extraction, multimodal localization, and multimodal data quality assessment of buildings in Bavaria. We plan to update the dataset with each major re-publication of the upstream data sources to foster change detection research in the future. The BBD is available at https://doi.org/10.14459/2023mp1709451.

The ACM SIGSPATIAL Cup 2023 proposed the challenge to identify and map supraglacial lakes in Greenland in satellite imagery. The peculiarities of supraglacial lakes pose a hard problem for semantic segmentation and object detection tasks because the definition of a lake is ill-fitted to the inner workings of such approaches. For example, lakes are often covered by ice and snow and narrow streams can connect distinct lakes, which is not directly translatable to the semantic segmentation of water. It is also not well-posed for object detection, especially the identity relation - what is a lake, what is not (yet) a lake, and what are two lakes is challenging. In this context, we worked on adapting semantic segmentation using the Segment Anything Model and instance segmentation using Mask R-CNN to the setting. The latter ended up superior in our own evaluation and even got ranked second among all participants. We are proud that our approach has led to competitive performance. The source code is available from https://github.com/tum-bgd/GISCup23.