Abstract
Cars have become an essential mode of transportation for people all around the world, but unfortunately, traffic accidents remain a persistent problem. This paper proposes a solution to this issue using “deep learning” models and “image processing” algorithms, without the need for any additional sensors. The proposed system is designed to categorize images with 99.92% accuracy into two classes, “Night” and “Day”, using a specially designed filter. Based on the results, the system employs a deep learning model that achieves 87.66% accuracy for vehicle detection, 80.47% accuracy for pedestrian detection, and 88.80% accuracy for lane detection. The system provides feedback to the driver based on these results, thereby enhancing driver awareness and improving overall road safety.
Bariscan Kurtkaya
My research interests include generative models and 2D & 3D computer vision.