DFineNet: Ego-Motion Estimation and Depth Refinement from Sparse, Noisy Depth Input with RGB Guidance
This addresses depth estimation challenges for autonomous systems, offering a solution that works with cheaper sensors, but it appears incremental as it builds on existing depth refinement and completion methods.
The paper tackles the problem of depth estimation for autonomous vehicles by proposing an end-to-end learning algorithm that refines sparse, noisy depth input using RGB guidance, also producing camera pose as a byproduct; it shows strong performance on KITTI and superior handling of sparse, noisy depth on TUM datasets.
Depth estimation is an important capability for autonomous vehicles to understand and reconstruct 3D environments as well as avoid obstacles during the execution. Accurate depth sensors such as LiDARs are often heavy, expensive and can only provide sparse depth while lighter depth sensors such as stereo cameras are noiser in comparison. We propose an end-to-end learning algorithm that is capable of using sparse, noisy input depth for refinement and depth completion. Our model also produces the camera pose as a byproduct, making it a great solution for autonomous systems. We evaluate our approach on both indoor and outdoor datasets. Empirical results show that our method performs well on the KITTI~\cite{kitti_geiger2012we} dataset when compared to other competing methods, while having superior performance in dealing with sparse, noisy input depth on the TUM~\cite{sturm12iros} dataset.