Face alignment with an ensemble of regression tree
Introduction
Face alignment is a process that estimates the face¡¯s landmarks position or facial shape from face images, and it is widely used in research and commercial applications, including, object pose estimation, face recognition, 3D reconstruction and automatic face beautification. However, the efficiency and accuracy to estimate the face¡¯s landmarks position did not meet practical requirements until 2014. At that time, a milestone model [1] was invented to solve this problem directly from a sparse subset of pixel intensities, achieving super-realtime and high prediction accuracy performance on standard datasets. The new model was based on an ensemble of gradient boosting tree that optimizes the sum of landmarks position square error loss and appropriate priors of image data structure in feature selection. Your task is to implement the algorithms used in the paper and train the model in iBUG 300W dataset.
Data description: iBUG 300W
iBUG 300W dataset composes of 600 face images with corresponding coordinate annotations of 68 landmarks.
download link:https://ibug.doc.ic.ac.uk/resources/facial-point-annotations/ Requirements
1. You need to submit a pdf format report, in which you should clearly describe the method and explain the result.
2. The code should be uploaded to Github and the Github link should be provided in the report.
Reference
[1] Kazemi V, Sullivan J One millisecond face alignment with an ensemble of regression tree. In Proceedings of the IEEE conference on computer vision and pattern recognition, 2014.
Contact TA
Xiao Jiashun (jxiaoae@connect.ust.hk)