High Accuracy Face Geometry Capture using a Smartphone Video

  • 2020-03-19 05:46:58
  • Shubham Agrawal, Anuj Pahuja, Simon Lucey
  • 27

Abstract

What's the most accurate 3D model of your face you can obtain while sittingat your desk? We attempt to answer this question in our work. High fidelityface reconstructions have so far been limited to either studio settings orthrough expensive 3D scanners. On the other hand, unconstrained reconstructionmethods are typically limited by low-capacity models. Our method reconstructsaccurate face geometry of a subject using a video shot from a smartphone in anunconstrained environment. Our approach takes advantage of recent advances invisual SLAM, keypoint detection, and object detection to improve accuracy androbustness. By not being constrained to a model subspace, our reconstructedmeshes capture important details while being robust to noise and beingtopologically consistent. Our evaluations show that our method outperformscurrent single and multi-view baselines by a significant margin, both in termsof geometric accuracy and in capturing person-specific details important formaking realistic looking models.

 

Quick Read (beta)

loading the full paper ...