Pose Constraints for Consistent Self-supervised Monocular Depth and Ego-motion
This work addresses scale ambiguity and inconsistency issues in self-supervised monocular depth estimation, which is an incremental improvement for applications like robotics and autonomous driving.
The paper tackles the problem of temporally inconsistent depth maps in self-supervised monocular depth estimation by introducing temporal consistency losses to minimize pose inconsistencies, resulting in reduced depth inconsistencies and improved baseline performance for depth and ego-motion prediction.
Self-supervised monocular depth estimation approaches suffer not only from scale ambiguity but also infer temporally inconsistent depth maps w.r.t. scale. While disambiguating scale during training is not possible without some kind of ground truth supervision, having scale consistent depth predictions would make it possible to calculate scale once during inference as a post-processing step and use it over-time. With this as a goal, a set of temporal consistency losses that minimize pose inconsistencies over time are introduced. Evaluations show that introducing these constraints not only reduces depth inconsistencies but also improves the baseline performance of depth and ego-motion prediction.