A novel landmark detector system for multi resolution frontal faces
MetadataShow full item record
In this paper, we implement a facial landmarking system to improve the performance of landmark location accuracy for the tree-structured based facial detector proposed recently by Zhu and Ramanan. Our main objective is to overcome their limitation where very small faces could not be detected and landmarked. Furthermore, we also want to improve the landmarking accuracy and reduce false positive rate for facial images with various resolutions in one image. We achieve these aims by developing two separate tree-structured face models in an integrated system. The first one is the Multi Resolution (MR) models where it can detect faces on images of any resolution and further provide suitable number of landmarks. The second one is that we develop a Tree-structured Filter Model (TFM) which can reduce false positives quickly to avoid high processing time for multiple faces with different resolutions in one image. Finally, we combine these 2 models with Viola-Jones face detector to create a facial landmarking system. Our experiments show that our proposed models can detect small faces down to 30x30 pixels. Furthermore, our models can improve the landmarking accuracy as well as reduce false positive rates significantly.
Showing items related by title, author, creator and subject.
Liang, A.; Wang, C.; Liu, Wan-Quan; Li, L. (2016)In this paper, we propose a robust facial landmarking scheme for frontal faces which can be applied on both controlled and uncontrolled environ-ment. This scheme is based on improvement/extension of the tree-structured ...
Deep, dense and accurate 3D face correspondence for generating population specific deformable modelsGilani, S.; Mian, A.; Eastwood, Peter (2017)Â© 2017 Elsevier Ltd We present a multilinear algorithm to automatically establish dense point-to-point correspondence over an arbitrarily large number of population specific 3D faces across identities, facial expressions ...
Liang, Antoni; Liu, Wan-Quan; Li, Ling; Farid, M.; Le, V. (2014)In this paper, we aim to improve one of the current state-of-the-art models for facial components detection/localization. The objectives are to increase the amount of landmark points detected and improve the landmark ...