List All Projects List Graduate Projects List Senior Design Projects List REU Projects Other Undergraduate Projects
Project # 131
RoofVision: Good Neighbor
| Team Lead: | CONG THANH TRAN |
| Other Team Members: | Kasson Davis | Miwa Yoshida | Serah Jolly | Jeffery Aguirre | Miguel Oropeza |
| Faculty Advisor(s): | Shawn N Gieser |
| Department: | CSE |
| Project Type: | Senior Design |
| Title: | RoofVision: Good Neighbor |
| Abstract: | This project presents Roof Vision Good Neighbor, an AI driven roof inspection system developed to detect wind and hail damage efficiently and accurately. The solution integrates drone based image capture with a custom object detection model trained using Roboflow. High resolution aerial imagery is processed through a deep learning pipeline optimized with ONNX to enable efficient inference. The model is deployed on Google Cloud Platform to provide scalable and reliable cloud access. Performance evaluation demonstrates strong detection capability, achieving a mean average precision at IoU 0.50 of 85 percent for wind damage and 65 percent for hail damage. These results reflect robust identification of larger wind related defects and highlight ongoing opportunities to improve hail detection, which often presents with smaller and more subtle visual features. The system replaces manual inspections that are time consuming and inconsistent by automatically identifying damaged regions and generating structured visual reports. By reducing inspection time and operational cost while improving diagnostic consistency, Roof Vision Good Neighbor enables homeowners and service providers to assess roof conditions quickly without physical access. This project demonstrates the practical integration of drone technology, computer vision, and cloud deployment for scalable roof damage assessment. |
| Poster: | View Poster #131 (PDF) |
| Judges Scoring: | Submit Evaluation (judges only) |