In a long-term collaboration with Toyota Motor Europe, we have developed a system for tracking faces in videos by matching a 3D face model to the sequence of 2D images. The goal is to track the faces of drivers. This is challenging because of the considerable variation in head pose, lighting and expression exhibited, as well as occlusions caused by parts of the car or the driver's hands. The current system has been shown to be robust to these variations.
|
|
|
|
The following videos show the result of searching frames originally captured at 15 fps. Although they were generated offline they were simulated inside a real-time system capable of searching live video at up to 30 fps
 |
Facial Image Analysis |
 |
Tools for Understanding Facial Behaviour |
A.Caunce, C.Taylor and T.Cootes, "Using Detailed Independent 3D sub-models to improve facial feature localisation and pose estimation.", International Journal on Artificial Intelligence Tools, Vol.22, No.6, 2013 (here)
A.Caunce, T.F.Cootes and C.J.Taylor, "Using Detailed Independent 3D Sub-Models to Improve Facial Feature Localisation and Pose Estimation", Proc. International Symposium on Visual Computing, 2012, pp 398-408, (PDF)
A.Caunce, T.F.Cootes and C.J.Taylor, "Adding Facial Actions into 3D Model Search to Analyse Behaviour in an Unconstrained Environment", Proc. International Symposium on Visual Computing, 2010, pp 132-142, (PDF)
A.Caunce, T.F.Cootes and C.J.Taylor, "Improved 3D Model Search for Facial Feature Location and Pose Estimation in 2D images", Proc. British Machine Vision Conference 2010 (PDF),(Revised PDF)
A.Caunce, T.F.Cootes and C.J.Taylor, "Locating Facial Features and Pose Estimation Using a 3D Shape Model", Proc. International Symposium on Visual Computing, 2009,pp 750-761. (pdf),(Revised PDF)