3D Face Reconstruction from Multiple 2D Images

Photo of author
Written By Zach Johnson

AI and tech enthusiast with a background in machine learning.

Reconstructing 3D facial shapes from 2D images is critical to face analysis applications. However, it is difficult to get 3D ground truth shapes for face images because of the requirements for special sensors and constrained environment in data acquisition. In order to overcome the scarcity of 3D face datasets, a real-world, large-scale, high-precision multi-modal dataset, namely PixelAI-3DFace, is released to bridge the gap between 2D images and 3D shapes in this challenge.

Input: Left, frontal, and right side 2D facial images captured by a self-customized trinocular high-resolution camera.
Output: 106 3D facial landmarks.


  • 1st prize: USD$ 9,000
  • 2nd prize: USD$ 4,000
  • 3rd prize: USD$ 2,000

Schedule (Beijing, UTC+8)

  • April 1st, 10:00:00, 2021 – Training / validation set released
  • May 6th, 10:00:00, 2021 – Testing set released and submission opened
  • June 4th, 10:00:00, 2021 – Submission deadline
  • June 8th, 10:00:00, 2021 – Challenge winners notified
  • June 20th, 2021 – Winners present at CVPR 2021 Workshop

Please refer to the link for more details.

3DFace Dataset Details

Set Subjects Images 3D Scans 3D Meshes
Training Set 523 15,612 5,204 5,204
Validation Set 106 2,691 897 897
Testing Set 225 7,128 2,376 2,376

AI is evolving. Don't get left behind.

AI insights delivered straight to your inbox.

Please enable JavaScript in your browser to complete this form.