Im2avatar reconstructs colorful 3D model from a single-view image. It requires TensorFlow to run. It simultaneously recovers 3D shape and surface color from a single image. The application is based on an end-to-end trainable framework called Colorful Voxel Network (CVN). Conditioned on a single 2D input, CVN learns to decompose shape and surface color information of a 3D object into a 3D shape branch and a surface color branch, respectively.
- Yongbin Sun, Ziwei Liu, Yue Wang, Sanjay E Sarma, "Im2Avatar: Colorful 3D Reconstruction from a Single Image", arXiv preprint, 2018.