Lane Curvature Detection Using Bezier Curves and ENet Segmentation

This project implements Bezier curve-based lane detection techniques for accurate curvature estimation in autonomous driving scenarios. It benchmarks performance against ENet segmentation models and utilizes the TuSimple dataset to improve lane detection accuracy.

Features

• Bezier Curve Lane Detection: Models lane lines with cubic Bezier curves for smoother lane curvature estimation.
• Performance Benchmarking: Compared against ENet segmentation models to validate performance improvements.
• Dataset: Utilizes the TuSimple dataset, improving lane detection accuracy by 15%.
• Real-Time Processing: Optimized algorithms to ensure real-time performance for autonomous navigation.

Project Structure

Lane-Curvature-Detection/ ├── notebooks/ │ ├── Bezier_Curve_Synthetic_Dataset_Final.ipynb │ ├── Bezier_With_Metrics_TuSimple_Dataset.ipynb │ └── ENet_With_Metrics.ipynb ├── presentation/ │ └── EE243_Computer_Vision_Project_Daksh_Akhil.pptx └── README.md

How to Run

1. Clone the repository:

   git clone https://github.com/akhiljoshi7060/Lane-Curvature-Detection.git cd Lane-Curvature-Detection

2. Run the Jupyter notebooks:

jupyter notebook notebooks/Bezier_With_Metrics_TuSimple_Dataset.ipynb