Investigation of pavement crack detection based on deep learning method using weakly supervised instance segmentation framework

This paper presents efficient and cost-effective methods to identify pavement crack distress and thereby substantially increase pavement strength. Detecting the origin of this distress is the key to restoring pavement performance. To do that, a deep learning method is used to detect cracks based on the weakly supervised instance segmentation (WSIS) framework. A bounding box-level crack image data is preprocessed. Pseudo labels are generated by a region growing algorithm and a GrabCut algorithm. Another important contribution is a new dynamically balanced binary cross-entropy loss function. Results show that the WSIS framework reduces manual marking and has a high recognition accuracy of crack distress.

Automated summary

This paper presents efficient and cost-effective methods to identify pavement crack distress using deep learning and the weakly supervised instance segmentation framework, which reduces manual marking and achieves highly accurate recognition of crack distress.