Localizing Parts of Faces Using a Consensus of Exemplars

By Peter N. Belhumeur, David W. Jacobs, David J. Kriegman, Neeraj Kumar. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013

1,069 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/140438a77a771a8fb656b39a78ff488066eb6b50

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Do Convnets Learn Correspondence? (2014, background): “Most of the existing works on part localization or keypoint prediction focus on either facial landmark localization [11] or human pose estimation.”
Extensive Facial Landmark Localization with Coarse-to-Fine Convolutional Network Cascade (2013, methodology): “The images and annotations come from AFW, LFPW, HELEN, and IBUG [6, 1, 12, 7, 8].”
Constrained Local Neural Fields for Robust Facial Landmark Detection in the Wild (2013, methodology): “It was trained using LFPW [4] and Multi-PIE datasets.”
Face detection, pose estimation, and landmark localization in the wild (2012, background): “Recent work has focused on global spatial models built on top of local part detectors, sometimes known as Constrained Local Models (CLMs) [11, 34, 5].”
POOF: Part-Based One-vs.-One Features for Fine-Grained Categorization, Face Verification, and Attribute Estimation (2013, methodology): “We use part locations from the part detector of [2] on images cropped to the bounding boxes of the birds, allowing us to compare with previous work that uses the bounding boxes but not the part labels.”
Adaptive Wing Loss for Robust Face Alignment via Heatmap Regression (2019, methodology): “300W consists of the following subsets: LFPW [2], HELEN [32], AFW [72], XM2VTS [40] and an additional dataset with 135 images with large pose, occlusion and expressions called iBUG.”
Learning an appearance-based gaze estimator from one million synthesised images (2016, background): “2013] and LFPW+Helen [Belhumeur et al. 2011; Le et al. 2012], containing 135, 337, 600, and 554 images respectively.”
Dog Breed Classification Using Part Localization (2012, methodology): “Once extracted, these descriptors are then concatenated into a single 1024-dimensional feature vector for our SVM regressor.”
Adaptive 3D Face Reconstruction from Unconstrained Photo Collections (2017, background): “In recent years, the automatic detection of landmarks [11], [50], [51] has seen rapid improvement due to large labeled datasets such as LFPW [52] and 300-”
WarpNet: Weakly Supervised Matching for Single-View Reconstruction (2016, neutral): “The output grid, normalized to a range of [−1, 1]× [−1, 1], acts as the control points for computing a grid-based TPS transformation from I2 to I1.”
Supervised Descent Method and Its Applications to Face Alignment (2013, methodology): “We show how SDM improves state-of-the-art performance for facial feature detection in two “face in the wild” databases [26, 4] and demonstrate extremely good performance tracking faces in the YouTube celebrity database [20].”
The First Facial Landmark Tracking in-the-Wild Challenge: Benchmark and Results (2015, background): “Arguably, the most comprehensive such benchmark was ﬁrstly presented in [46] and then continued in [44] (so-called 300-”
Whole-Body Human Pose Estimation in the Wild (2020, background): “Second, the variations such as illumination, pose and scales in the existing human face [4,26,29,33,49,69], hand [56,13,13,34], and body datasets [3,2,62,31] are different, inevitably introducing dataset biases to the learned deep networks, thus hindering the development of algorithms to comprehensi”
Nonparametric Part Transfer for Fine-Grained Recognition (2014, background): “The paper of [21] builds on [2] for part detection.”
DenseReg: Fully Convolutional Dense Shape Regression In-the-Wild (2016, methodology): “Our training set consists of the LFPW train-set [2], Helen trainset [25] and AFW [49], thus 3148 images that are captured under completely unconstrained conditions and exhibit large variations in pose, expression, illumination, age, etc.”
Face Alignment by Explicit Shape Regression (2012, methodology): “The method in [1] uses expensive local landmark detectors (SIFT+SVM) and it takes more than 10 seconds6 to run 29 detectors over the entire image.”
Neural best-buddies (2018, background): “Most of the existing works on part localization or keypoint prediction focus on either facial landmark localization [Belhumeur et al. 2013; Kowalski et al. 2017] or human pose estimation [Gkioxari et al.”
(2013) 300 Faces in-the-Wild Challenge: the first facial landmark localization challenge. In: 2013 IEEE International (2017, methodology): “To facilitate the training procedure, landmark annotations for LFPW [2], HELEN [9], AFW [19] and XM2VTS [13] became available from the 300-W challenge’s website 1.”