Feng Liu bio photo

Feng Liu

  G. Scholar Github E-Mail

To Provide 3D Faces For 2D Face Databases (To be released)

 3DFaceProvider 

We aim to provide photo-realistic 3D face models for the famous 2D face databases, such as Multi-PIE, LFW, 300W, etc. We believe that such 3D face geometry models will advance the 3D-based techniques towards applications in real-world scenarios.

Multi-Dim Face Database

 Multi-Dim Face Database 

We build a new benchmark database that includes both 2D and 3D face data, both still images and surveillance videos, both acted and spontaneous data, and display miscellaneous variations in pose, illumination, expression, resolution, occlusion and motion blur. The multi-dimensional face database is called Multi-Dim. We believe that such a database will advance the 3D-based unconstrained 2D face recognition and related techniques towards applications in real-world scenarios.

3D Face Reconstrcution and Recognition

3D Face Reconstrcution and Recognition 

Given single or multiple 2D face images of arbitrary poses and expressions, we aim to reconstruct 3D face models and apply them to enhance unconstrained face recognition accuracy.

3D-assisted Face Alignment

3D-assisted Face Alignment 

Existing studies tackle the problems of face alignment and 3D face reconstruction separately. However, these two problems are chicken-and-egg problems. On one hand, 2D face images are projections of 3D faces onto the 2D plane. Knowing a 3D face and a 3D-to-2D mapping function, it is easy to compute the visibility and position of 2D landmarks. On the other hand, the landmarks provide rich information about facial geometry, which is the basis of 3D face reconstruction.