During the summer break of 2022, an NTU civil engineering team guided by Prof. Jacob J Lin participated in the 3rd Annual VIMS / 1st VIMS-IAARC Joint Datathon Competition 2022. This datathon was jointly organized by the American Society of Civil Engineers (ASCE) Computing Division’s Visualization, Information Modeling, and Simulation (VIMS) Committee and the International Association for Automation and Robotics in Construction’s (IAARC) Workshops and Summer Schools Committee from June 24 – July 5, 2022. The team consisted of a Ph.D. student, Mr. Aritra Pal, and four undergraduate students: Ms. Ting-Hsuan Lin, Mr. Cheng Huai-En, Mr. Lin Zi-Jun, and Mr. Jen-Yu Lu.
With the increasing awareness of sustainable development, reusing and recycling waste is gaining importance. Research has shown that a lot of construction and demolition waste materials can be effectively used in new construction. Can such material be automatically identified through computer vision and segregated using robots? The Datathon’s objective was to discover what is possible through computer vision. Therefore, the goal was to identify recyclable objects such as PVC pipes and rebars within construction debris through image processing and artificial intelligence.
The NTU team, called NTUCV, proposed a deep-learning-based image segmentation approach to address the above-stated challenge. Within a week, they labeled more than 1200 images with polygon masks and trained three image segmentation models to detect five types of construction demolition wastes: brick, cementitious debris, PVC, rebar, and wires. They submitted their interim report to the organizing committee on June 30. Based on the interim report, 20 teams were selected to participate in the final competition on July 5. On the final competition day, the judges provided a test dataset on the spot, and the team had to run their model and show test results within 5 minutes.
Congratulations to the team for winning the second position. Other winners were from the University of Wisconsin-Madison, USA (First) and Western Virginia University, USA (Third).