UROP Project

Privacy-Preserving Edge AI for Automated Demolition-Site Monitoring

edge computing, computer vision, Wi-Fi sensing networks, privacy-preserving AI, demolition automation

Research Mentor: Dr. Xin Liu,
Department, College, Affiliation: Department of Computer Science, Florida State University, Arts and Sciences
Contact Email: xliu15@fsu.edu
Research Assistant Supervisor (if different from mentor): Dr. Roshan Panahi
Research Assistant Supervisor Email: rpanahi@eng.famu.fsu.edu
Faculty Collaborators: Dr. Juyeong Choi
Faculty Collaborators Email: jchoi@eng.famu.fsu.edu

Looking for Research Assistants: Yes
Number of Research Assistants: 1
Relevant Majors: Computer Science, Data Science, Artificial Intelligence, Electrical/Computer Engineering, or Civil Engineering. Students from other majors with relevant programming or data-analysis experience are welcome.
Project Location: On FSU Main Campus
Research Assistant Transportation Required:
Remote or In-person: Partially Remote
Approximate Weekly Hours: 5-10 hours per week, Flexible schedule (Combination of business and outside of business. TBD between student and research mentor.)
Roundtable Times and Zoom Link:

• Day: Monday, August 31
  Start Time: 3:00
  End Time: 5:30
  Zoom Link: https://fsu.zoom.us/my/xinliucs

Project Description

Construction and demolition operations generate large amounts of waste, yet improving equipment productivity and material recovery requires reliable data from real jobsites. Video monitoring can provide this information, but contractors are often reluctant to share recordings because they may contain faces, equipment identifiers, accidents, or other sensitive events. This interdisciplinary project will develop a privacy-preserving edge-AI system that collects video from demolition equipment over a local Wi-Fi network, processes the data on an on-site gateway, and retains only the operational information needed for analysis. The system combines rugged vehicle-side cameras and long-range Wi-Fi links with event-triggered recording, detection and redaction of personally identifiable information, and computer-vision models that recognize equipment activities such as crushing, scooping, swinging, dumping, and loading. The research will evaluate whether privacy-filtered video still preserves enough information to estimate equipment cycle times accurately. The long-term goal is to make large-scale, privacy-aware data collection practical for demolition contractors and to support safer, more efficient, and more sustainable demolition planning.

Research Tasks: The research assistant will: (1) review literature on edge computing, privacy-preserving video analytics, object detection, and construction-equipment activity recognition; (2) help organize and document an egocentric/exocentric demolition-video dataset; (3) annotate selected frames or short clips for workers, equipment, faces/identifiers, and activity classes; (4) run and compare baseline computer-vision models for detection, tracking, redaction, and activity recognition; (5) evaluate model outputs using metrics such as precision, recall, F1-score, mAP, and cycle-time error; (6) assist with controlled testing of cameras, Wi-Fi links, and the on-site gateway; (7) maintain reproducible experiment notes and summarize findings in figures and short reports; and (8) contribute to the abstract and poster for the FSU Undergraduate Research Symposium. Site visits, if offered, will be optional and conducted only with required safety training and supervision.

Skills that research assistant(s) may need: Required: curiosity, reliability, willingness to learn, and basic experience with programming or quantitative analysis. Recommended: Python; introductory machine learning or computer vision; Linux/Git; data labeling; statistics; networking or wireless systems; and clear technical writing. Prior construction or demolition experience is not required. Training will be provided for project-specific tools, responsible handling of video data, and research documentation.