Overview / Usage

We would like a Movidius NCS in order to run a convolutional neural network for obstacle detection. Our team, RoboJackets, participates in the annual Intelligent Ground Vehicle Competition. This competition revolves around autonomously navigating an outdoor obstacle course by avoiding obstacles and driving to GPS waypoints. Classical line detection algorithms struggle in our applications due to the noisy and inconsistent nature of our competition environment. Our team is looking at moving our lane detection algorithms to a convolution neural network in order to infer boundaries based on segmentation and classification of drivable space. The competition revolves around limited onboard computing and sensing and this product will support machine learning approach on our limited power and computing resources. We wish to implement a robust convolutional neural network-based model that will allow us to interpret images at a rate that is commensurate with autonomous driving. All of our labeled data, network, and code will be entirely open source and publicly available in our Github repository.