Paper presented at IEEE ICIP 2022 in Bordeaux

1 August 2022

Bridges are an important part of the infrastructure worldwide but many bridges are reaching critical ages. Hence, digitized inspections including automated damage recognition can save time and raise inspection quality.

This research is interdisciplinary work with Johannes Flotzinger and Prof. Braml from the Institute for Structural Engineering at UniBw M. Together with Philipp J. Rösch and Prof. Oswald from VIS the authors provide the building-inspection-toolkit and evaluated the two multi-target datasets which are currently available in this domain. They run extensive hyperparameter tuning with three transfer learning strategies on three modern CNNs. Their results acts as new baseline for the datasets MCDS (Hüthwohl et al., 2019) and CODEBRIM (Mundt et al., 2019). The work is accepted at IEEE International Conference on Image Processing (ICIP 2022) and will be presented in Bordeaux.

Building Inspection Toolkit: Unified Evaluation and Strong Baselines for Bridge Damage Recognition

Johannes Flotzinger, Philipp J. Rösch, Norbert Oswald, Thomas Braml

[arXiv]