CVPR Daily - 2019 Thursday

23 DAILY CVPR Thursday Posters - Wednesday Top, Panna Felsen, Angjoo Kanazawa and Jason Y. Zhang presenting their poster Learning 3D Human Dynamics From Video . Angjoo told us how to reconstruct smooth 3D mesh humans from video without post processing and predict 3D humans motion from a single image. Below, Angela Dai presenting poster Scan2Mesh: From Unstructured Range Scans to 3D Meshes . Angela told us that scan2mesh shows that we can generate actual 3D meshes (indexed face sets) through graph-based interpretations -- this kind of mesh generation is important towards eventually achieving CAD-like meshes.