Impact of eye detection error on face recognition performance

Dutta, Abhishek; Günther, Manuel; Shafey, Laurent El; Marcel, Sébastien; Veldhuis, Raymond; Spreeuwers, Luuk

doi:10.1049/iet-bmt.2014.0037

Cited by 9 publications

(6 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We selected these parameters to be well outside of the error range of a reasonable (frontal face) eye detector. Alignment errors of these magnitudes have been shown to highly influence the performance of many traditional face recognition algorithms [34].…”

Section: Handling Mis-aligned Imagesmentioning

confidence: 99%

MOON: A Mixed Objective Optimization Network for the Recognition of Facial Attributes

Rudd

Günther

Boult

2016

Computer Vision – ECCV 2016

Self Cite

153

194

View full text Add to dashboard Cite

Abstract. Attribute recognition, particularly facial, extracts many labels for each image. While some multi-task vision problems can be decomposed into separate tasks and stages, e.g., training independent models for each task, for a growing set of problems joint optimization across all tasks has been shown to improve performance. We show that for deep convolutional neural network (DCNN) facial attribute extraction, multi-task optimization is better. Unfortunately, it can be difficult to apply joint optimization to DCNNs when training data is imbalanced, and re-balancing multi-label data directly is structurally infeasible, since adding/removing data to balance one label will change the sampling of the other labels. This paper addresses the multi-label imbalance problem by introducing a novel mixed objective optimization network (MOON) with a loss function that mixes multiple task objectives with domain adaptive re-weighting of propagated loss. Experiments demonstrate that not only does MOON advance the state of the art in facial attribute recognition, but it also outperforms independently trained DCNNs using the same data. When using facial attributes for the LFW face recognition task, we show that our balanced (domain adapted) network outperforms the unbalanced trained network.

show abstract

Section: Handling Mis-aligned Imagesmentioning

confidence: 99%

MOON: A Mixed Objective Optimization Network for the Recognition of Facial Attributes

Rudd

Günther

Boult

2016

Computer Vision – ECCV 2016

Self Cite

153

194

View full text Add to dashboard Cite

show abstract

“…As shown in [5,22], face recognition works better on aligned faces. Therefore, for most face processing tasks, the first step is to align the face based on facial landmarks in order to extract features from it [32,33].…”

Section: Introductionmentioning

confidence: 98%

AFFACT: Alignment-free facial attribute classification technique

Günther

Rozsa

Boult

2017

2017 IEEE International Joint Conference on Biometrics (IJCB)

View full text Add to dashboard Cite

Facial attributes are soft-biometrics that allow limiting the search space, e.g., by rejecting identities with nonmatching facial characteristics such as nose sizes or eyebrow shapes. In this paper, we investigate how the latest versions of deep convolutional neural networks, ResNets, perform on the facial attribute classification task. We test two loss functions: the sigmoid cross-entropy loss and the Euclidean loss, and find that for classification performance there is little difference between these two. Using an ensemble of three ResNets, we obtain the new state-of-the-art facial attribute classification error of 8.00 % on the aligned images of the CelebA dataset. More significantly, we introduce the Alignment-Free Facial Attribute Classification Technique (AFFACT), a data augmentation technique that allows a network to classify facial attributes without requiring alignment beyond detected face bounding boxes. To our best knowledge, we are the first to report similar accuracy when using only the detected bounding boxes -rather than requiring alignment based on automatically detected facial landmarks -and who can improve classification accuracy with rotating and scaling test images. We show that this approach outperforms the CelebA baseline on unaligned images with a relative improvement of 36.8 %.

show abstract

“…In this paper, we employed the VGG-face CNN model, with two major contributions:  The proposed method involves no preprocessing steps, i.e., images are only resized to the standard size of VGG-model input without any form of registration. Registration steps, usually used in the literature [13,[17][18][19], may not only involve alignment errors [20,21], but also make the method less robust, time consuming, and sensitive to the success of the registration step.  The extracted features using the transfer learning VGG-model are further optimized using a sophisticated GA in order to select the most relevant features to AIFR.…”

Section: Introductionmentioning

confidence: 99%

Optimization of deep learning features for age-invariant face recognition

Moustafa¹,

Elnakib²,

Areed³

2020

IJECE

View full text Add to dashboard Cite

This paper presents a methodology for Age-Invariant Face Recognition (AIFR), based on the optimization of deep learning features. The proposed method extracts deep learning features using transfer deep learning, extracted from the unprocessed face images. To optimize the extracted features, a Genetic Algorithm (GA) procedure is designed in order to select the most relevant features to the problem of identifying a person based on his/her facial images over different ages. For classification, K-Nearest Neighbor (KNN) classifiers with different distance metrics are investigated, i.e., Correlation, Euclidian, Cosine, and Manhattan distance metrics. Experimental results using a Manhattan distance KNN classifier achieves the best Rank-1 recognition rate of 86.2% and 96% on the standard FGNET and MORPH datasets, respectively. Compared to the state-of-the-art methods, our proposed method needs no preprocessing stages. In addition, the experiments show its privilege over other related methods.

show abstract

Impact of eye detection error on face recognition performance

Cited by 9 publications

References 22 publications

MOON: A Mixed Objective Optimization Network for the Recognition of Facial Attributes

MOON: A Mixed Objective Optimization Network for the Recognition of Facial Attributes

AFFACT: Alignment-free facial attribute classification technique

Optimization of deep learning features for age-invariant face recognition

Contact Info

Product

Resources

About