Mind the Gap - A Deep Learning Analysis of Pothole Detection скачать в хорошем качестве

Mind the Gap - A Deep Learning Analysis of Pothole Detection

In this project, we trained state-of-the-art object detection models for pothole detection and evaluated their performance under self-synthesized motion blur noise. We analyzed the impact of motion blur on detection accuracy and improved system robustness by training on augmented data. Additionally, we introduced a classification system that categorizes detected potholes into three severity levels based on their risk to pedestrians and vehicles. We then retrained the best-performing model on this enhanced dataset to improve its effectiveness. This project was conducted as part of our Deep Learning course (#046211) at Technion.