Sparse2Dense: From direct sparse odometry to dense 3D reconstruction
This work addresses the problem of accurate positioning and high-quality depth reconstruction in monocular SLAM for robotics or AR/VR applications, representing an incremental advance by integrating learned depth and normals into an existing framework.
The paper tackled dense 3D reconstruction from monocular SLAM by using a deep learning framework that maps sparse odometry to dense models via learned surface normals, resulting in significant improvements in visual tracking and depth prediction compared to state-of-the-art methods.
In this paper, we proposed a new deep learning based dense monocular SLAM method. Compared to existing methods, the proposed framework constructs a dense 3D model via a sparse to dense mapping using learned surface normals. With single view learned depth estimation as prior for monocular visual odometry, we obtain both accurate positioning and high quality depth reconstruction. The depth and normal are predicted by a single network trained in a tightly coupled manner.Experimental results show that our method significantly improves the performance of visual tracking and depth prediction in comparison to the state-of-the-art in deep monocular dense SLAM.