Abstract:
In recent years, deep learning approaches have shown superior performance in car lane detection compared to traditional methods. However, the computational demands of deep learning algorithms often hinder their real-time application in autonomous vehicles. To address this issue, this study proposes a lane detection technique that combines the power of deep learning with real-time requirements. A lightweight convolutional neural network model is employed as a feature extractor, trained on a dataset of 16 x 64 pixel tiny image patches. Fast inference is achieved using a non-overlapping sliding window technique, and lane boundaries are modeled by fitting a polynomial to the predictions. The proposed technique is evaluated on the KITTI and Caltech datasets, demonstrating satisfactory performance. Additionally, we integrate the detector into our autonomous vehicle's localization and planning system, achieving a real-time processing speed of 28 frames per second on a CPU with an image resolution of 768 x 1024, meeting the requirements for self-driving cars.
At Smart AI Technologies, we specialize in providing comprehensive support for final year engineering projects based on IEEE papers. As one of the leading project centers in Bangalore, we offer the following components for your project:
Complete documentation support to ensure well-documented project work.
Implementation of working hardware/software in the students' environment.
Conducting classes and guidance tailored to the project's requirements.
By choosing our project center in Bangalore, you can ensure that your final year engineering project on car lane detection using machine learning will be among the best IEEE projects for computer science. Contact us now to avail our services and receive expert guidance throughout the project development process.