Live Demonstration: Unsupervised Event-Based Learning of Optical Flow, Depth and Egomotion

We propose a demo of our work, Unsupervised Event-based Learning of Optical Flow, Depth and Egomotion, which will also appear at CVPR 2019. Our demo consists of a CNN which takes as input events from a DAVIS-346b event camera, represented as a discretized event volume, and predicts optical flow for each pixel in the image. Due to the generalization abilities of our network, we are able to predict accurate optical flow for a very wide range of scenes, including for very fast motions and challenging lighting.

我们提议展示我们的研究成果——基于无监督事件的光流、深度和自我运动学习，该成果也将在2019年计算机视觉与模式识别会议（CVPR）上亮相。我们的演示包含一个卷积神经网络（CNN），它以DAVIS - 346b事件相机输出的事件作为输入，这些事件以离散化的事件体表示，该网络可预测图像中每个像素的光流。由于我们网络的泛化能力，我们能够针对非常广泛的场景预测出精确的光流，包括极快速的运动场景以及具有挑战性的光照场景。