AN APPROACH FOR DETECTING BUILDING DEFECTS USING DEEP LEARNING CONVOLUTION NEURAL NETWORKS

Customers are increasingly looking for fast and efficient ways to frequently check and inspect the condition of their buildings so essential repairs and maintenance can be done in a timely manner. Act quickly before they become too dangerous and costly. Traditional methods for this type of work commonly comprise engaging inspection officers to undertake a condition assessment which involves a lengthy site inspection to produce a systematic recording of the building defects, including cost estimates of immediate and project the long-term costs of renewal, repair and maintenance of the building. Current asset condition assessment procedures are extensively time-consuming and expensive and pose health and safety threats to inspection officers, particularly at height and roof levels which are difficult to access. This project aims toward an automated detection and localization of key building defects, e.g., moss, cracks, deterioration and stains, from images using Convolution Neural Networks. The proposed model was built on a pre-trained CNN classifier of VGG-16, with class activation mapping (CAM) using a squeeze net for object localization. The challenges and limitations of the model in real-life applications have been identified. The proposed neural network model has proved that it can accurately detect and localize building defects.