On November 28, 2025, Christian Benz successfully defended his dissertation entitled “Image-based Crack Detection for Structural Inspection” at the Bauhaus-Universität Weimar. The thesis convincingly demonstrates how image acquisition, training data for deep learning, and algorithms for neural networks influence detection performance under difficult conditions. Configured and trained with the OMNICRACK30k dataset, nnU-CRACKNET delivers promising results in the detection of cracks on different surface materials. The new clIoU metric allows for slight positional uncertainties to be taken into account when comparing linear crack structures on a pixel-by-pixel basis. The spatial localization of cracks in structures facilitates the assessment of their severity. For this purpose, the semi-synthetic and real data sets CRACKENSEMBLES and CRACKSTRUCTURES were generated. The crack detection in 2.5D, referred to as ENSTRECT, is realized with the help of a reconstructed 3D point cloud. The proposed fusion scheme successfully detects surface cracks in practical applications. However, the performance of crack segmentation depends heavily on image quality, and “in-the-wild” conditions continue to pose a challenge.