Facial alignment by using sparse initialization and random forest

Liew, Chun Fui; Yokoya, Naoto; Yairi, Takehisa

doi:10.1109/icip.2014.7025057

Cited by 1 publication

(6 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this section, we examine six different shape initialization methods, such as the conventional mean shape initialization [5], sparse initialization [26], multiple initialization [12], and two-stage initialization. In the conventional mean shape initialization method, the mean shape of all training data is used as the initialized shape.…”

Section: Comparison Of Shape Initialization Methodsmentioning

confidence: 99%

“…In the conventional mean shape initialization method, the mean shape of all training data is used as the initialized shape. In the sparse initialization method, both PC-shapes and AT-shapes are computed from the training data and then the optimized initialized shape will be selected by maximizing an objective function [26]. In multiple initialization method, random training shapes are used as initialized shapes and the median of fitting results are taken as the final results.…”

Section: Comparison Of Shape Initialization Methodsmentioning

confidence: 99%

“…While being effective most of the time, it would most likely fail if the face being aligned has a true shape that is very different or far away from the mean shape. On the other hand, sparse initialization method has been proposed to solve this problem [26]. Figure 3 illustrates the sparse initialization method with one test image.…”

Section: Shape Initializationmentioning

confidence: 99%

“…We compare two types of shape initialization in our experiments: the mean shape initialization and sparse shape initialization [26]. Mean shape initialization uses the mean shape computed from the training data as the initialized shape.…”

Section: Shape Initializationmentioning

confidence: 99%

“…A process flow of sparse initialization [26]. By optimizing an objective function inspired by sparse model, sparse initialization can identify the best candidate and enhance alignment performance.…”

Section: Figmentioning

confidence: 99%

See 4 more Smart Citations

Robust Face Alignment with Random Forest: Analysis of Initialization, Landmarks Regression, and Shape Regularization Methods

Liew

Yairi

2016

IEICE Trans. Inf. & Syst.

Self Cite

View full text Add to dashboard Cite

SUMMARYRandom forest regressor has recently been proposed as a local landmark estimator in the face alignment problem. It has been shown that random forest regressor can achieve accurate, fast, and robust performance when coupled with a global face-shape regularizer. In this paper, we extend this approach and propose a new Local Forest Classification and Regression (LFCR) framework in order to handle face images with large yaw angles. Specifically, the LFCR has an additional classification step prior to the regression step. Our experiment results show that this additional classification step is useful in rejecting outliers prior to the regression step, thus improving the face alignment results. We also analyze each system component through detailed experiments. In addition to the selection of feature descriptors and several important tuning parameters of the random forest regressor, we examine different initialization and shape regularization processes. We compare our best outcomes to the state-of-the-art system and show that our method outperforms other parametric shape-fitting approaches. key words: face alignment, shape model fitting, random forest, point distribution model, sparse representation model

show abstract

Section: Comparison Of Shape Initialization Methodsmentioning

confidence: 99%

Section: Comparison Of Shape Initialization Methodsmentioning

confidence: 99%

Section: Shape Initializationmentioning

confidence: 99%

Section: Shape Initializationmentioning

confidence: 99%

Section: Figmentioning

confidence: 99%

See 3 more Smart Citations