Constrained Local Neural Fields for Robust Facial Landmark Detection in the Wild

By Tadas Baltrusaitis, Peter Robinson, Louis-philippe Morency. 2013 IEEE International Conference on Computer Vision Workshops, 2013

438 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/35dc5d70ed3c74c484e7d5f00a83660cb65c4074

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

Improving social skills in children with ASD using a long-term, in-home social robot (2018, methodology): “With RGB camera stream input, the system estimated and tracked head poses and orientations by using Constrained Local Model face tracking and landmark detection algorithms (34) and approximated attentional targets according to the estimated head poses and orientations.”
RT-GENE: Real-Time Eye Gaze Estimation in Natural Environments (2018, background): “Comparison of various landmark detectors [3,25] on the original images (with eyetracking glasses), images where the eyetracking glasses are ﬁlled with a uniform color (the mean color of the image), and inpainted images as proposed in our method.”
Cross-dataset learning and person-specific normalisation for automatic Action Unit detection (2015, methodology): “We use Constrained Local Neural Field (CLNF) [2] facial landmark detector and tracker for face tracking and to extract geometry based features (explained in Section IV-D).”
OpenFace 2.0: Facial Behavior Analysis Toolkit (2018, methodology): “In order to estimate eye gaze, we use a Constrained Local Neural Field (CLNF) landmark detector [9], [70] to detect eyelids, iris, and the pupil.”
High-Resolution Representations for Labeling Pixels and Regions (2019, neutral): “[3] - 19.”
Every Smile is Unique: Landmark-Guided Diverse Smile Generation (2018, methodology): “are extracted using [4], and 64 × 64 binary images are created from them.”
Mnemonic Descent Method: A Recurrent Process Applied for End-to-End Face Alignment (2016, background): “t=0 W ho H (t) 2 F (5) 4 We omit the bias terms for brevity of notation. where H (t) = [h (t) 1 , . . . , h (t) n ] ∈ R u×n represents the matrix of all states corresponding to each of the n images at time t, and θ all the parameters of the model.”
A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems (2019, background): “Recent advances in deep neural networks, and their successful application for highdimensional data regression in image recognition [11, 12, 13] and natural language processing [14, 15], have stimulated research on deep-learning-based surrogate modeling for high-dimensional nonlinear systems.”
Image and Vision Computing (2016, positive): “[41] propose a probabilistic patch expert technique that learns non-linear and spatial relationships between the pixels and the probability of a landmark being aligned.”
Eye Tracking for Everyone (2016, methodology): “First, while we designed the iTracker network to be robust to poor-quality eye detections, we use tighter crops (of size 80× 80) produced by facial landmark eye detections [1] for the smaller network.”
(2013) 300 Faces in-the-Wild Challenge: the first facial landmark localization challenge. In: 2013 IEEE International (2017, background): “[1] proposed a probabilistic patch expert (landmark detector) that can learn non-linear and spatial relationships between the input pixels and the probability of a landmark being aligned.”
Rendering of Eyes for Eye-Shape Registration and Gaze Estimation (2015, methodology): “First, our system CLNF Synth (Mdn = 0.0110) trained on only 10 participants in four lighting conditions results in very similar performance to a system trained on unconstrained in-the-wild images, CLNF 300-”
Learning Pose-Aware Models for Pose-Invariant Face Recognition in the Wild (2019, methodology): “By improving face-preprocessing using a newer and more robust version of the CLNF face detector [58], [64] and by using unaligned images, we were able to obtain state-ofthe-art performance on PIPA with an accuracy of 70.”
OpenFace: An open source facial behavior analysis toolkit (2016, methodology): “[8].”
Do We Really Need to Collect Millions of Faces for Effective Face Recognition? (2016, methodology): “Face detections: Our pipeline uses the facial landmark detector of [2] for head pose estimation and alignment.”