FCN for Segmentation Applied To Structural Defect Detection
, Screening Eagle Technologies
, Screening Eagle Dreamlab
, Screening Eagle Dreamlab
, Screening Eagle Dreamlab
The poster shows a fully synthetically trained system using NVIDIA GPUs for dataset generation, training, and inference that helps identify structural damage and defects, such as cracks, spalling, efflorescence, biological growth, etc., in buildings and infrastructure. The generation of the synthetic dataset uses Unity game engine, in communication via JSON to the configuration parameters exposed by Substance, to parametrically generate the dataset at lightning speed with a state-of-the-art infrastructure deployment in Kubernetes that manages the GPUs to produce it. The impact that this project had in the construction and inspection industry is enormous, given the previous lack of precise data. It now gives full precise detection of the above-mentioned defects.