A mini-review of face alignment - classical method versus deep learning approach / Nabilah Hamzah, Fadhlan Hafizhelmi Kamaru Zaman and Nooritawati Md Tahir

Face alignment is one of the vital research areas to be explored specifically face tasks like face recognition, face verification, face reconstruction, and facial expression analysis. Hence, the need for robust face alignment is still in demand. Numerous classic methods have used the 2D image for the detection of facial landmarks but this task is challenging due to several reasons, for instance, large poses, semi-frontal images, and facial expression. Abundant techniques have been established to mitigate all these challenges but there are far from being solved. Hence this mini-review discussed the face alignment methods based on the classic method to the state-of-the-art that includes the 2D-face alignment along with the 3D-face alignment approach. Based on the review done, the 3D model could combat large poses, facial expressions, and semi-frontal images however some of the facial landmarks are not visible and stack together for occluded face images. Hence, this will be the research area to be explored further in ensuring robustness and better accuracy in the face alignment area.