This project focuses on developing a road segmentation model tailored for autonomous driving applications. The model uses the U-Net architecture to segment roads from stereo images effectively, enhancing the navigational capabilities of autonomous vehicles.
- Architecture: U-Net
- Purpose: Road segmentation
- Features: Integration of depth maps with road ground truth to enhance segmentation accuracy.
- Dataset Used: KITTI Vision Benchmark Suite
- Description: A widely recognized dataset comprising high-resolution stereo image pairs designed for autonomous driving research, capturing detailed urban environment data.
- Reference: KITTI Vision Benchmark Suite
For a detailed discussion of the project's methodology, implementation, and results, please refer to the report.pdf
in this repository. This document provides a comprehensive analysis of the model development, training procedures, evaluation metrics, and the project's challenges.