Dyna-DM: Dynamic Object-aware Self-supervised Monocular Depth Maps
This addresses depth estimation for robotics and autonomous driving, offering a more efficient approach with competitive results.
The paper tackles self-supervised monocular depth estimation by improving the learning process rather than increasing model complexity, achieving state-of-the-art performance with a 29% reduction in GPU memory usage.
Self-supervised monocular depth estimation has been a subject of intense study in recent years, because of its applications in robotics and autonomous driving. Much of the recent work focuses on improving depth estimation by increasing architecture complexity. This paper shows that state-of-the-art performance can also be achieved by improving the learning process rather than increasing model complexity. More specifically, we propose (i) disregarding small potentially dynamic objects when training, and (ii) employing an appearance-based approach to separately estimate object pose for truly dynamic objects. We demonstrate that these simplifications reduce GPU memory usage by 29% and result in qualitatively and quantitatively improved depth maps. The code is available at https://github.com/kieran514/Dyna-DM.