27-02-2017, 10:07 AM
Pose-Invariant Face Alignment
Face alignment is a process of applying a monitored learned model to an image of the face and estimating the locations of a set of facial reference points, such as corners of the eye, corners of the mouth, etc. Face alignment is a dominant module in the pipeline of most facial analysis algorithms, typically after face detection and prior extraction and classification of features. Therefore, it is an enabling capability with a multitude of applications, such as facial recognition, expression recognition, face detection, etc. Despite continuous improvement in alignment accuracy, alignment of the face remains a very difficult problem, due to the non-frontal Face Pose, low image quality, occlusion, etc. Among all the challenges, we identify the alignment of invariant faces as deserving substantial research efforts, for a number of reasons.
Click here to download the source code
Face alignment is a process of applying a monitored learned model to an image of the face and estimating the locations of a set of facial reference points, such as corners of the eye, corners of the mouth, etc. Face alignment is a dominant module in the pipeline of most facial analysis algorithms, typically after face detection and prior extraction and classification of features. Therefore, it is an enabling capability with a multitude of applications, such as facial recognition, expression recognition, face detection, etc. Despite continuous improvement in alignment accuracy, alignment of the face remains a very difficult problem, due to the non-frontal Face Pose, low image quality, occlusion, etc. Among all the challenges, we identify the alignment of invariant faces as deserving substantial research efforts, for a number of reasons.
Click here to download the source code