Self-supervised Object Motion and Depth Estimation from Video
This work addresses the challenge of scale ambiguity in motion prediction for autonomous driving applications, though it appears incremental by building on existing self-supervised methods.
The paper tackles the problem of estimating individual object motion and monocular depth from video using a self-supervised framework, achieving improved performance in 3D scene flow prediction on the KITTI dataset without external annotations.
We present a self-supervised learning framework to estimate the individual object motion and monocular depth from video. We model the object motion as a 6 degree-of-freedom rigid-body transformation. The instance segmentation mask is leveraged to introduce the information of object. Compared with methods which predict dense optical flow map to model the motion, our approach significantly reduces the number of values to be estimated. Our system eliminates the scale ambiguity of motion prediction through imposing a novel geometric constraint loss term. Experiments on KITTI driving dataset demonstrate our system is capable to capture the object motion without external annotation. Our system outperforms previous self-supervised approaches in terms of 3D scene flow prediction, and contribute to the disparity prediction in dynamic area.