Efficient Monocular Depth Estimation for Edge Computing Platforms

Estimating depth is necessary to understand and navigate the environment surrounding us. Over the years, many active sensors have been developed to measure depth, but they are expensive and require additional space for mounting. A cheaper alternative is estimating depth from a single RGB image taken by an ordinary monocular camera, which can be placed inside the smartphone. However, state-of-the-art depth estimation algorithms are based on complex deep neural networks that are too slow for real-time inference on mobile platforms which can be mounted, for instance, on a micro aerial vehicle. That fact is a barrier to the further development of monocular depth estimation. In this paper, we address this problem. We utilize recent advancements in the architecture of lightweight networks to reduce complexity. We propose a novel lightweight network design with competitive accuracy and significant complexity reduction compared to existing approaches. Our methodology indicates that it is possible to achieve inference speeds accelerated by an order of magnitude on NVIDIA Jetson Nano and, at the same time, preserve the comparable accuracy on the KITTI Odometry dataset in comparison with the current state-of-the-art algorithms.