Academic project
Automatic Detection of Concrete Cracks Using Computer Vision
Investigation of U-Net semantic segmentation for concrete crack detection, evaluated with CrackForest.
Inspecting cracks in concrete structures can require access to buildings, bridges, and overpasses in hard-to-reach locations. The project proposes to investigate convolutional neural networks for automated identification of damaged regions in images.
The Lattes description specifies U-Net for semantic segmentation, the public CrackForest dataset, and the Dice metric for evaluation. It also mentions Mogai Tecnologia de Informação's HammerHead computer vision system and a proposed adaptation for deployment on a drone. These are the project's stated aims and methods; completion of the integration has not been verified.
Team listed in Lattes: Jefferson Oliveira Andrade, Karin Satie Komati (project lead), Kelly Assis de Souza Gazolli, Daniel Ribeiro Trindade. Lattes lists this project as “ongoing”; its current status has not been independently checked.