One millisecond face alignment with an ensemble of regression trees

Kazemi, Vahid; Sullivan, Josephine

doi:10.1109/cvpr.2014.241

Cited by 2,319 publications

(1,564 citation statements)

References 16 publications

Supporting

Mentioning

1,440

Contrasting

Unclassified

Order By: Relevance

“…Burgos-Artizzu et al [12] integrate part visibility term into landmarks and presents interpolated shape-indexed features to tackle with occlusions and high shape variances. Kazemi et al [32] estimate facial landmarks by learning an ensemble of regression trees (ERT) directly from a sparse subset of pixel intensities. Their ERT achieves millisecond performance and can handle partial or uncertain labels, but the correlation of shape parameters is little taken into account.…”

Section: Cascaded Regression To Face Alignmentmentioning

confidence: 99%

See 1 more Smart Citation

Sign-Correlation Partition Based on Global Supervised Descent Method for Face Alignment

Zhang

Liu

Yang

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. Face alignment is an essential task for facial performance capture and expression analysis. As a complex nonlinear problem in computer vision, face alignment across poses is still not studied well. Although the state-of-the-art Supervised Descent Method (SDM) has shown good performance, it learns conflict descent direction in the whole complex space due to various poses and expressions. Global SDM has been presented to deal with this case by domain partition in feature and shape PCA spaces for face tracking and pose estimation. However, it is not suitable for the face alignment problem due to unknown ground truth shapes. In this paper we propose a sign-correlation subspace method for the domain partition of global SDM. In our method only one reduced low dimensional subspace is enough for domain partition, thus adjusting the global SDM efficiently for face alignment. Unlike previous methods, we analyze the sign correlation between features and shapes, and project both of them into a mutual sign-correlation subspace. Each pair of projected shape and feature keep sign consistent in each dimension of the subspace, so that each hyperoctant holds the condition that one general descent exists. Then a set of general descent directions are learned from the samples in different hyperoctants. Our sign-correlation partition method is validated in the public face datasets, which includes a range of poses. It indicates that our methods can reveal their latent relationships to poses. The comparison with state-of-the-art methods for face alignment demonstrates that our method outperforms them especially in uncontrolled conditions with various poses, while keeping comparable speed.

show abstract

Section: Cascaded Regression To Face Alignmentmentioning

confidence: 99%

“…We evaluate the proposed sign-correlation partition SDM method on the challenging 300W dataset, and compare it with state-of-the-art methods ESR [10], SDM [11], ERT [32], and LBF [33]. As mentioned above, there are 68 labeled landmarks in this dataset.…”

Section: Comparison Of Face Alignmentmentioning

confidence: 99%

Sign-Correlation Partition Based on Global Supervised Descent Method for Face Alignment

Zhang

Liu

Yang

et al. 2017

Lecture Notes in Computer Science

View full text Add to dashboard Cite

show abstract

“…All frames are cropped to the size of 224 × 224. Then 68 facial landmark points are detected for all frames in the dataset as in [18]. In Fig.…”

Section: A Implementation Detailsmentioning

confidence: 99%

Changes in Facial Expression as Biometric: A Database and Benchmarks of Identification

Haamer

Kulkarni

Imanpour

et al. 2018

2018 13th IEEE International Conference on Automatic Face &Amp; Gesture Recognition (FG 2018)

View full text Add to dashboard Cite

Abstract-Facial dynamics can be considered as unique signatures for discrimination between people. These have started to become important topic since many devices have the possibility of unlocking using face recognition or verification. In this work, we evaluate the efficacy of the transition frames of video in emotion as compared to the peak emotion frames for identification. For experiments with transition frames we extract features from each frame of the video from a fine-tuned VGG-Face Convolutional Neural Network (CNN) and geometric features from facial landmark points. To model the temporal context of the transition frames we train a Long-Short Term Memory (LSTM) on the geometric and the CNN features. Furthermore, we employ two fusion strategies: first, an early fusion, in which the geometric and the CNN features are stacked and fed to the LSTM. Second, a late fusion, in which the prediction of the LSTMs, trained independently on the two features, are stacked and used with a Support Vector Machine (SVM). Experimental results show that the late fusion strategy gives the best results and the transition frames give better identification results as compared to the peak emotion frames.

show abstract

“…Since the human face is non-rigid and large nonlinear deformations occur in extreme expressions, [4] used a Haar-like feature based Adaboost face detector [33] to initialize the face location, then adopted a Supervised Descent Method (SDM) to refine the locations of facial landmarks. [8] presented a state-of-the-art method for facial landmark detection with super-real time performance, which used an ensemble of regression trees to estimate the positions of facial landmark accurately from a sparse subset of pixel intensities. The ensemble of regression trees was learned based on gradient boosting.…”

Section: Related Workmentioning

confidence: 99%

“…There have been many recent achievements in related research sub-areas such as facial landmark localization [3][4] [5][6] [7] [8], tracking and recognition [9] [10]. Realistic facial expression synthesis is useful for affective computing, human computer interaction [11] [12], realistic computer animation [13] and facial surgery planning [14][15], etc.…”

Section: Introductionmentioning

confidence: 99%

Facial expression transfer method based on frequency analysis

Wei

Tian

Maybank

et al. 2016

Pattern Recognition

View full text Add to dashboard Cite

Abstract. We propose a novel expression transfer method based on an analysis of the frequency of multi-expression facial images. We locate the facial features automatically and describe the shape deformations between a neutral expression and non-neutral expressions. The subtle expression changes are important visual clues to distinguish different expressions. These changes are more salient in the frequency domain than in the image domain. We extract the subtle local expression deformations for the source subject, coded in the wavelet decomposition. This information about expressions is transferred to a target subject. The resulting synthesized image preserves both the facial appearance of the target subject and the expression details of the source subject.This method is extended to dynamic expression transfer to allow a more precise interpretation of facial expressions.Experiments on Japanese Female Facial Expression (JAFFE), the extended Cohn-Kanade (CK+) and PIE facial expression databases show the superiority of our method over the state-of-the-art method.

show abstract

One millisecond face alignment with an ensemble of regression trees

Cited by 2,319 publications

References 16 publications

Sign-Correlation Partition Based on Global Supervised Descent Method for Face Alignment

Sign-Correlation Partition Based on Global Supervised Descent Method for Face Alignment

Changes in Facial Expression as Biometric: A Database and Benchmarks of Identification

Facial expression transfer method based on frequency analysis

Contact Info

Product

Resources

About