Multi-task feature learning-based improved supervised descent method for facial landmark detection

Facial landmark detection has played an important role in many face understanding tasks, such as face verification, facial expression recognition, age estimation et al. Model initialization and feature extraction are crucial in supervised landmark detection. Mismatching caused by detector error and discrepant initialization is very common in these existing methods. To solve this problem, we have proposed a new method called multi-task feature learning-based improved supervised descent method (MtFL-iSDM) for the robust facial landmark localization. In this new method, firstly, a fast detection will be processed to locate the eyes and mouth, and the initialization model will adapt to the real location according to fast facial points detection. Secondly, multi-task feature learning is adopted on our improved supervised descent method model to achieve a better performance. Experiments on four benchmark databases show that our method achieves state-of-the-art performance.