3D Face Alignment Without Correspondences

Sánta, Zsolt; Kató, Zoltán

doi:10.1007/978-3-319-48881-3_36

Cited by 13 publications

(27 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some other methods can reconstruct 3D faces without 3D shape basis, while still rely on a 3D facial template. [36,37,38,17,39] can produce a 3D structure by warping the shape of a reference 3D model. [16] also reconstruct the 3D shape of faces by learning a 3D Thin Plate Spline(TPS) warping function via a deep network which warps a generic 3D model to a subject specific 3D shape.…”

Section: D Face Reconstructionmentioning

confidence: 99%

“…However, 2D landmarks location only regresses visible points on faces, which leads to difficulties and is limited to describe face shape when the pose of faces is large. Recent works then research the 3D facial alignment, which begins with fitting a 3DMM [47,11,48] or registering a 3D facial template [39,49] with a 2D facial image. Obviously, 3D reconstruction methods based on model can easily complete the task of 3D face alignment by selecting x,y coordinates of landmarks vertices in reconstructed geometry.…”

Section: Face Alignmentmentioning

confidence: 99%

See 1 more Smart Citation

Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network

Feng

Shao

et al. 2018

Lecture Notes in Computer Science

651

583

View full text Add to dashboard Cite

We propose a straightforward method that simultaneously reconstructs the 3D facial structure and provides dense alignment. To achieve this, we design a 2D representation called UV position map which records the 3D shape of a complete face in UV space, then train a simple Convolutional Neural Network to regress it from a single 2D image. We also integrate a weight mask into the loss function during training to improve the performance of the network. Our method does not rely on any prior face model, and can reconstruct full facial geometry along with semantic meaning. Meanwhile, our network is very light-weighted and spends only 9.8ms to process an image, which is extremely faster than previous works. Experiments on multiple challenging datasets show that our method surpasses other state-of-the-art methods on both reconstruction and alignment tasks by a large margin. Code is available at https://github.com/YadiraF/PRNet.

show abstract

Section: D Face Reconstructionmentioning

confidence: 99%

Section: Face Alignmentmentioning

confidence: 99%

Joint 3D Face Reconstruction and Dense Alignment with Position Map Regression Network

Feng

Shao

et al. 2018

Lecture Notes in Computer Science

651

583

View full text Add to dashboard Cite

show abstract

“…Let us formulate now the alignment problem: Given a pair of template and observation objects denoted by F t ⊂ R 3 and F o ⊂ R 3 , respectively, we are looking for the aligning transformation ϕ such that for all x ∈ F t there exists a y ∈ F o satisfying the so-called identity relation [Sánta and Kato, 2018…”

Section: Problem Statementmentioning

confidence: 99%

“…In the first approach, the control points have This method is also known as the Farthest Point Sampling [102]. Note that, this approach places the control points on the surface only [Sánta and Kato, 2018] [Sánta and Kato, 2016a].…”

Section: Control Point Placementmentioning

confidence: 99%

“…, ℓ over one triangular element will be O(M 3 ), where M is the maximal degree of polynomials from the {ω i } ℓ i=1 set. Now, let us focus on the approximate calculation of the integrals [Sánta and Kato, 2018] [Sánta and Kato, 2016a]. Using the barycentric parametrization from Equation 3.35we linearly interpolate the {ω i } functions using the vertices of each triangle o = (A, B, C)…”

Section: )mentioning

confidence: 99%

See 1 more Smart Citation

Non-Rigid Registration of Visual Objects

Sánta

Self Cite

View full text Add to dashboard Cite

Sample results on the Bosphorus dataset. In each row, we show the template, the observation, the warped template and the landmark locations, respectively. In the last column, green denotes the ground truth position and red is the estimated location. While the proposed approach achieved good results near nose and mouth areas, the highest errors are near the eyebrows.. .. .. .. . 53 LIST OF FIGURES v 3.20 Comparison between the proposed, the GMMREG [5] and the CPD [6] algorithms. On the left diagram, the D GT measure is presented, while on the right we show the running times of each algorithm. The GMMREG slightly outperformed the proposed approach and the CPD algorithm in terms of alignment accuracy, but the proposed approach has the best running time.. .. .. .. .

show abstract