A method and a system for facial landmark detection based on multi-task

Abstract

The present application disclosed a method and system for detecting facial landmarks of a face image. The method may comprise extracting multiple feature maps from at least one facial region of the face image and/or the whole face image; generating a shared facial feature vector from the extracted multiple feature maps; and predicting facial landmark locations of the face image from the generated shared facial feature vector. With the present method and system, the facial landmark detection can be optimized together with heterogeneous but subtly related task, so that the detection robustness can be improved through multi-task learning.

Description

technical field [0001] The present application relates to face alignment, in particular, to methods and systems for face landmark detection. Background technique [0002] Facial keypoint detection is fundamental to many face analysis tasks, such as face attribute inference, face verification, and face recognition, but has been hampered by the problems of light occlusion and pose variation. [0003] Accurate facial keypoint detection can be performed using a cascaded CNN (Convolutional Neural Network), where the face is pre-partitioned into different parts, and each part is processed by a separate deep CNN. The resulting outputs are then averaged and passed to separate cascaded layers to process each face keypoint separately. [0004] Furthermore, facial keypoint detection is not an independent problem, and its estimation can be affected by many heterogeneous but subtly correlated factors. For example, when a child is smiling, his or her mouth is opened wide. Effectively d...

AI Technical Summary

Problems solved by technology

[0005] However, different tasks will inherently differ in learning difficulty and have different convergence rates

Furthermore, some t

Images