The present study proposes a novel machine learning architecture that can specifically reconstruct civil structural system from a hand-sketch image domain to a finite-element mathematical domain, referred to as an Image-based Structural Analysis. The proposed Image-based Structural Analysis facilitates learning progress of civil engineering students in understanding the structural response under certain loading conditions. From practical point of view, it has a great potential to speed up the design process in civil engineering projects. The study will employ some searching algorithms to segment different components of structural objects in an image and train a supervised deep convolutional neural network to classify the detected objects from the segmentation for further system reconstruction. The study requires participating students to investigate various efficient deep convolutional neural network that developed in recent years and several searching algorithms for fast object segmentation.
In this project, students will be charged to conduct the following two tasks:
(1) Explore and investigate various deep convolutional neural network architectures for image-based structural analysis;
(2) Proposes suitable deep learning architecture for segment objects from a hand-drawn image.
The project is opened to all levels of students from engineering and science.
(1) Explore and investigate various deep learning methods
(2) Develop self-learning and self-motivating attributes