The is the facial tracking component developed for the FROG project (http://www.frogrobot.eu/). The programme takes as an video clip and provides an output, for each frame, of the 66 facial landmarks as depicted in Figure 1, as well as the pitch, yaw and roll angles defining the face pose estimation. The module is based on a novel discriminative regression based approach for the Constrained Local Models (CLMs) framework, referred to as the Discriminative Response Map Fitting (DRMF) method, which shows impressive performance in the generic face fitting scenario.
Figure 1: The 66 points mark-up provided by the tracker.
For detailed description of the method, please refer to:
Robust Discriminative Response Map Fitting with Constrained Local Models. A. Asthana, S. Zafeiriou, S. Cheng, M. Pantic. 2013 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2013). Portland, Oregon, USA, June 2013.
Please note the programme requires CUDA 2.0-compatiable graphics card to run.
The download link: