Gate Detection for Micro Aerial Vehicles using a Single Shot Detector

Abstract

Object detection is a common approach in the area that works with images as robotics and image processing. The conventional techniques to identify an object is to apply segmentation, edge detector or feature extraction. Actually, deep learning has been used to solve multiple tasks including classification, segmentation and detection using classic CNN as VGG16, YOLO , AlexNet and more. We present an approach to detect gate in a race path to racers drones using a Single Shot Detector Network (SSD) modified based on the optimiser SSD7. In comparison with other architectures of networks, we test our SSD in three different environments to compare the prediction time, average fps and the confidence obtained, achievement a proper detection of the gate and high confidence.