Real-time combined 2D+3D active appearance models

By Jing Xiao, Simon Baker, Iain Matthews, Takeo Kanade. Proceedings of the 2004 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2004. CVPR 2004., 2004

525 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/1a8dd02400bd0816a135fad36e111da99b66e091

How researchers see this paper

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

Citations in the corpus

2D vs. 3D Deformable Face Models: Representational Power, Construction, and Real-Time Fitting (2007, methodology): “In particular, 3D models such as 3D Morphable Models (Blanz and Vetter, 1999; Blanz et al., 2003) and 2D + 3D Active Appearance Models (Xiao et al., 2004) can be constructed from images collected with the same cameras that will be used in the final application.”
Estimating 3D shape and texture using pixel intensity, edges, specular highlights, texture constraints and a prior (2005, methodology): “[14] presented an algorithm estimating the 3D shape of a face from a single image by fitting a 2D Active Appearance Model (AAM) and constraining the resulting shape to be a valid 2D projection of a 3D shape modeled by a 3DMM.”
Measuring Facial Action (2008, methodology): “Like the cylindrical head model, the person-speciWc head model is robust to occlusion and runs at frame rate (30 frames per second) or faster (Gross et al. 2004; Xiao et al. 2004)”
Generic vs. person specific active appearance models (2005, background): “Using the argument in [12], 15×6 images should be sufficient to model the shape of all neutral faces across pose.”
Real-time expression transfer for facial reenactment (2015, methodology): “A prominent example is active appearance models (AAM) [Cootes et al. 2001] which are used to determine the parameters of a 3D PCA model while only using 2D features [Xiao et al. 2004].”
The First Facial Landmark Tracking in-the-Wild Challenge: Benchmark and Results (2015, background): “Rigid and non-rigid tracking of faces and facial features has been a very popular topic of research over the past twenty years [21, 29, 20, 17, 25, 37, 43, 23, 61, 19, 49, 14, 31, 34, 62, 63, 64, 10, 32, 40, 51, 39, 50, 35, 24, 26, 53, 56, 57, 54].”
Face Alignment in Full Pose Range: A 3D Total Solution (2015, background): “Considering the recent achievements in 3D face reconstruction which can construct a 3D face from 2D landmarks [1, 44, 55, 20], we assume that a 3D model can be accurately fitted if sufficient 2D landmarks are provided.”
Continuous sign language recognition: Towards large vocabulary statistical recognition systems handling multiple signers (2015, background): “[72] Xiao, J.”
3D Constrained Local Model for rigid and non-rigid facial tracking (2012, methodology): “Treating the locations of the true landmarks as hidden variables, they can be marginalised out of the likelihood that the landmarks are aligned: p(li|xi, I,Z) = ∑ yi∈Ψi p(li|yi, I,Z)p(yi|xi), (8) where p(yi|xi) = N (yi;xi, ρI), with ρ denoting the variance of the noise on landmark locations arising ”
Painful data: The UNBC-McMaster shoulder pain expression archive database (2011, methodology): “occurred, we used the 3D parameters from the AAM to estimate the pitch, yaw and roll [22].”
Automatic feature localisation with constrained local models (2008, methodology): “[16] have developed hybrid 2D–3D versions of the AAM and applied them to human faces.”
Sparse Representation for 3D Shape Estimation: A Convex Relaxation Approach (2015, methodology): “bility in 2D, such that we can get rid of the bilinear form in (6), which is a necessary step towards a convex formulation. The model in (9) is equivalent to the afﬁne-shape model in literature [42], [43], which uses an augmented linear space to represent the shape variation in 2D caused by both int”
Head Pose Estimation in Computer Vision: A Survey (2009, unknown): “The methods that track the head in video sequences using flexible models and tracking methods [14], [81], [82], [94], [124], [140], [147] report a significantly lower error than systems that estimate pose in individual images.”
State of the Art on Monocular 3D Face Reconstruction, Tracking, and Applications (2018, methodology): “Some methods are holistic, using the whole face region to localize the facial landmarks like Active Appearance Models (AAM) [CTCG95,CET01,MB04, XBMK04].”