Team
Daniel Steven, Rahul John, Antra Nayudu, Harman Brar, Ederuvie Udusegbe
Categories
Artificial Intelligence, Deep Learning, Python
Overview
This project focuses on accurate curb segmentation in an urban setting using the YOLOv8 segmentation model. A custom dataset of curb images was manually annotated to capture various curb types and conditions. The model was trained on the annotated dataset and evaluated using metrics such as mAP and IoU. This specific model was chosen for its speed and precision, balancing real time inference capability with high accuracy, which is paramount in for autonomous vehicles operating in real time. The results demonstrate strong performance in detecting and segmenting curbs under varying lighting and occlusion conditions.
Curb Detection Demo
Curb detection is an important task for small-scale autonomous vehicles operating in public. This presentation showcases our quick and efficient solution, perfect for autonomous vehicles.
