Published Paper Details:

The goal is to demonstrate a proof of concept for using transfer learning to address computer vision challenges in real time. This project involves developing a neural network-based vehicle control and driver assistance system focused on path detection and obstacle detection, enabling the bot to "learn" optimal responses for tasks like obstacle avoidance and path navigation toward a target destination. Initially, the bot will navigate along a given path, with the resulting performance data serving as the learning foundation. This learned information will then be applied in a more complex environment featuring obstacles and traffic elements such as signals and STOP signs. The objective is for the bot to traverse the course, avoid obstacles, and continuously learn as it moves. Transfer learning is advantageous in this context, as it reduces computational resource demands and time, achieving high accuracy efficiently. This problem is particularly challenging, combining areas like computer vision and machine learning, and requiring real-time processing capabilities. Python will be used for implementing the learning algorithms due to its wide availability and extensive library support, which makes it user-friendly and effective. The bot itself will be compact, no larger than a slice of bread, with a mounted microprocessor and microcontroller that enable it to methodically navigate the specified path. Powered by motors, the bot will autonomously and gradually work its way through the environment, showcasing the potential of transfer learning in real-time computer vision applications.