Introducing Pinscreen' latest #virtualavatar digitization technology from a single input photo, published at CVPR 2021. After year's of development, our brilliant team of researchers and engineers have come up with an approach that can generate so called #normalizedavatars using an approach based on StyleGAN and Perceptual refinement. The idea is that a 3D parametric model of a neutral avatar can be reconstructed robustly from extremely challenging input photos (random expressions, lighting conditions, head poses) without directly fitting to the input photo, but through a deep neural network that can extract and synthesize the likeness from the input. This technology will be incorporated soon into our next Pinscreen #avatarsdk with several additions including #superresolution and full body integration and improved #hairdigitization. For early access contact us at info@pinscreen.com. #cvpr2021#pinscreen#ucberkeley#virtualavatar#styleGAN Here is the paper: https://arxiv.org/abs/2106.11423And here is our paper video: https://youtu.be/V9r_jOiGX84 Stay tuned and visit: www.pinscreen.com to learn more.