Depth Distillation: Unsupervised Metric Depth Estimation for UAVs by Finding Consensus Between Kinematics, Optical Flow and Deep Learning
Estimating precise metric depth is an essential task for UAV navigation. Nevertheless, it is very difficult to do unsupervised learning without access to odometry. At the same time, depth recovery from kinematics and optical flow is mathematically precise, but less numerically stable and robust, especially in the focus of expansion areas. We propose a model that combines the analytical approach with deep learning, into a single formulation for metric depth estimation, that is both fast and accurate. The two pathways form a robust ensemble, which provides supervision to a single deep net that distills in this manner the consensus between scene geometry, pose, kinematics, camera intrinsics and the input RGB. The distilled net has low runtime and memory costs, being suitable for embedded devices. We validate our results against an off-the-shelf SfM-based solution. We also introduce a new real-world dataset of almost 20 minutes of continuous UAV flight, on which we demonstrate superior accuracy and capabilities to previous deep learning and classical approaches.