Wide-angle vision and sensing using artificial intelligence, machine learning and neural networks -- phase 2

该项目的工业合作伙伴ImmerVision专门从事新型广角全景光学和软件技术的开发。在寻求为客户提供额外价值的过程中，他们希望在产品中加入额外的“智能”。在这个项目的第一阶段，我们使用深度学习技术来估计广角图像的深度。在这个项目的第二阶段，我们建议使用深度学习来补偿视频会议背景下广角图像的失真。事实上，在这些图像中，人脸经常出现失真和不自然。其目标是利用数据驱动的方法来自动解开广角镜头拍摄的视频中的人脸。

The industrial partner in this project, ImmerVision, specializes in the development of novel wide-anglepanoramic optical and software technologies. In their quest to provide additional value to their customers, theywish to include additional "intelligence" to their products. In the first stage of this project, we used deeplearning techniques to estimate depth from wide angle images. In the second stage of this project, we proposeto use deep learning to compensate for distortions in wide angle images in the context of video conferencing.Indeed, faces often appear distorted and un-natural in these kinds of images. The goal is to exploit data-drivenapproaches to automatically unwrap faces in videos captured with wide-angle lenses.