Nonlinear Hierarchical Part-Based Regression for Unconstrained Face Alignment

Authors: Xiang Yu, Zhe Lin, Shaoting Zhang, Dimitris N. Metaxas

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on several challenging faces-in-the-wild datasets demonstrate the consistently better accuracy of our method, when compared to the state-of-the-art. We evaluate our method on six challenging benchmarks, i.e., Labeled Faces in the Wild (LFW) [Huang et al., 2007a], Labeled Facial Parts in the Wild (LFPW) [Belhumeur et al., 2011], Helen [Le et al., 2012], Annotated Faces-in-the-Wild (AFW) [Zhu and Ramanan, 2012], i Bug [Sagonas et al., 2013] and Caltech Occluded Faces in the Wild (COFW) [Burgos-Artizzu et al., 2013].
Researcher Affiliation Collaboration NEC Laboratories America, Media Analytics Adobe Research ]University of North Carolina at Charlotte Rutgers, The State University of New Jersey
Pseudocode Yes Algorithm 1 The two-stage regression algorithm.
Open Source Code No The paper states: 'The codes are provided by the authors from internet.' and 'RCPR provides its training code in which the annotation can be defined by input data.' This refers to the code of *other* methods, not the authors' own code for the proposed method (HPR).
Open Datasets Yes We evaluate our method on six challenging benchmarks, i.e., Labeled Faces in the Wild (LFW) [Huang et al., 2007a], Labeled Facial Parts in the Wild (LFPW) [Belhumeur et al., 2011], Helen [Le et al., 2012], Annotated Faces-in-the-Wild (AFW) [Zhu and Ramanan, 2012], i Bug [Sagonas et al., 2013] and Caltech Occluded Faces in the Wild (COFW) [Burgos-Artizzu et al., 2013].
Dataset Splits No The paper mentions training on LFPW and Helen but does not provide specific percentages or counts for training, validation, or test splits. It states: 'The training databases for HPR, RCPR and Co R are the same, which are LFPW and Helen.'
Hardware Specification Yes The average runtime on a 640 by 480 image is around 0.3s in Matlab with a dual core i7 3.4GHz CPU.
Software Dependencies No The paper mentions 'Matlab' but does not specify a version number or any other software dependencies with their versions.
Experiment Setup Yes Typically the number of iterations is fixed to 4 or 5. We set the threshold angles to be 22.5 and 22.5 . When the training samples are not sufficient, we augment the initialization by rotation and random perturbation of global translation. Meanwhile, to prevent overfitting, denoting s = s s0, we modify Eq. 3 by adding the regularization terms, which is Eq. 10.