Wide-angle vision and sensing using artificial intelligence, machine learning and neural networks

点击翻译按钮获取中文摘要

The industrial partner in this project, ImmerVision, specializes in the development of novel wide-angle**panoramic optical and software technologies. In their quest to provide additional value to their customers, they**wish to include additional "intelligence" to their products. Of particular interest, they wish to allow the**estimation of depth from a single camera, and are thus naturally turning towards deep learning approaches,**which have shown significant progress at this task in the recent. Unfortunately, all of these techniques have**been developed for cameras with more "traditional" fields of view (i.e. typically ranging from 60 to 90 degrees**of horizontal field of view), and cannot be applied "as is" for wide-angle images which typically exhibit**significant distortions, especially at the edges of the image. In addition, they cannot be re-trained directly on**wide-angle imagery, because there exists no dataset where both wide-angle images and corresponding depth**information is available. In this project, our objective is to develop a novel deep learning algorithm to estimate**depth from a monocular wide-angle image of an outdoor scene.