Deep learning-based smart vision for building and construction application

The current study was not created with a specific goal in view, which addresses many applications in the field of building and construction. It provides a variety of goals that must be addressed immediately for construction applications. This means that we approach the situation from different angles. The present paper approaches the title in two distinct ways. With regard to "building," we concentrated on damage detection (crack and spall). At the same time, for "construction site application," we mainly depend on worker safety. Since this study is a novel concept with two distinct domains, with one solution. We design an advanced deep learning strategy to address both of the study's objectives. Most previous studies specifically deal with this type of application with various approaches, but the present study gives one solution to address both the objectives. Guaranteeing stable structure of buildings and worker safety are important in the quickly developing field of construction management. By creating an edge-enhanced multi-drone system that makes use of advanced deep learning algorithms, this work provides a novel solution. The chief objectives of this research are to: (1) detect crack damage at building sites; and (2) Focused on employees' compliance by using drone technology. According to this, the proposed model combines two effective strengths called YOLOv3 and Bayesian optimization, an Intelligent Deep Learning (INDEED) model which helps to meet the dual challenges was handled by the study. The system's integration of edge computing principles guarantees real-time processing and decision-making abilities, and providing quick response to abnormalities occurred. Due to this combined ability, the model able to solve both damage detection and safety enhancement procedures in single hand. The evaluation of the study is conducted through two distinct datasets: Crack damage detection was assessed using CSIR-CEERI, Pilani. And the workers safety procedures were evaluated using the UAE based local site samples. The proposed model uses the hardware environment of Jetson-TX2. A multi-drone system, which helps to capture the surroundings of construction and buildings stability. The experiment proves the suggested model efficacy with two distinct validations.