Anytime Stereo Image Depth Estimation on Mobile Devices

TLDR

Stereo depth estimation for robotics must produce accurate disparity maps in real time while operating under strict computational limits, yet existing state‑of‑the‑art methods either sacrifice accuracy for speed or require excessive parameters. The study introduces an anytime disparity prediction method that adapts computation to achieve a balance between speed and accuracy. The method uses an end‑to‑end neural network that estimates depth in progressive stages, allowing the model to be queried at any point to output its current best disparity map. On a Jetson TX2, the model achieves 10–35 FPS on 1242×375 images with only slight error increase while employing two orders of magnitude fewer parameters than the most competitive baseline. Source code is available at https://github.com/mileyan/AnyNet.

Abstract

Many applications of stereo depth estimation in robotics require the generation of accurate disparity maps in real time under significant computational constraints. Current state-of-the-art algorithms force a choice between either generating accurate mappings at a slow pace, or quickly generating inaccurate ones, and additionally these methods typically require far too many parameters to be usable on power- or memory-constrained devices. Motivated by these shortcomings, we propose a novel approach for disparity prediction in the anytime setting. In contrast to prior work, our end-to-end learned approach can trade off computation and accuracy at inference time. Depth estimation is performed in stages, during which the model can be queried at any time to output its current best estimate. Our final model can process 1242×375 resolution images within a range of 10-35 FPS on an NVIDIA Jetson TX2 module with only marginal increases in error - using two orders of magnitude fewer parameters than the most competitive baseline. The source code is available at https://github.com/mileyan/AnyNet.