My research interest is at the intersection of computer vision, computer graphics, and machine learning areas, especially interested in reconstructing the 3D geometry and appearance of scenes (or objects). Below is the list of selected papers.
We complete an input depth image of sparse measurements effectively and efficiently by constructing a cost volume tailored to single-view depth completion task
We create the RSBlur, a novel dataset with real blurred image sequences and the corresponding sharp ones. We show that network training with realistic blurred images improves deblurring performance on real deblurred images.
We propose a novel framework that simultaneously reconstructs the geometry and texture map of a scene in real time. Thanks to the effective framework design and our image warping field estimation, we consequently can reconstruct a sharp textured 3D model in real time.