Effective training of convolutional neural networks for face-based gender and age prediction

Antipov, Grigory; Baccouche, Moez; Berrani, Sid-Ahmed; Dugelay, Jean‐Luc

doi:10.1016/j.patcog.2017.06.031

Cited by 121 publications

(79 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Amongst the different types of deep learning architectures, convolutional neural networks (CNN) have been proven to be very effective for human demographics estimation due to their proficiency at extracting precise details from images. Such studies include age estimation [13], [14], [15] and gender classification [16], [17], [18] . Niu et al [19] obtain an error of 3.28 years using ordinal regression CNNs and random splits of the MORPH-II dataset where 80% of the images are used for training and 20% are used for testing.…”

Section: Related Workmentioning

confidence: 99%

“…With a random split of 80% for training and 20% for testing on MORPH-II, it achieves a MAE of 2.68 with additional fine-tuning on IMDB-WIKI dataset before fine-tuning on MORPH-II dataset. Later, Antipov et al [20] extend the work from [13] and consider the problems of selection of optimal CNN architecture and training strategies. They conclude that Label Distribution Age Encoding (LDAE) [21] is an optimal way for the target encoding to train a CNN for an age estimation task.…”

Section: Related Workmentioning

confidence: 99%

“…A variety of network designs and training techniques are explored. We consider dynamic LDAE, which outperforms the static LDAE considered in [20]. A gender-specified hierarchical age model is proposed in this study.…”

Section: Related Workmentioning

confidence: 99%

“…LDAE is a method of encoding age that has proven more effective than simple one-hot encoding [16]. LDAE recognizes that people age differently, so it helps to view a person's age, denoted by A, as a small scope of potential ages rather than just a binary truth.…”

Section: Label Distribution Age Encoding (Ldae)mentioning

confidence: 99%

See 3 more Smart Citations

Transfer Learning with Deep CNNs for Gender Recognition and Age Estimation

Smith

Chen

2018

2018 IEEE International Conference on Big Data (Big Data)

View full text Add to dashboard Cite

In this project, competition-winning deep neural networks with pretrained weights are used for image-based gender recognition and age estimation. Transfer learning is explored using both VGG19 and VGGFace pretrained models by testing the effects of changes in various design schemes and training parameters in order to improve prediction accuracy. Training techniques such as input standardization, data augmentation, and label distribution age encoding are compared. Finally, a hierarchy of deep CNNs is tested that first classifies subjects by gender, and then uses separate male and female age models to predict age. A gender recognition accuracy of 98.7% and an MAE of 4.1 years is achieved. This paper shows that, with proper training techniques, good results can be obtained by retasking existing convolutional filters towards a new purpose.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Label Distribution Age Encoding (Ldae)mentioning

confidence: 99%

See 2 more Smart Citations

Transfer Learning with Deep CNNs for Gender Recognition and Age Estimation

Smith

Chen

2018

2018 IEEE International Conference on Big Data (Big Data)

View full text Add to dashboard Cite

show abstract

“…It has been used for a variety of applications in face recognition and demographic analysis. In particular, the MORPH-II dataset is widely utilized in research on gender [2] and race classification [3], age estimation [4], [5], [6], [7], [8], and age synthesis [9]. However, there are some challenges with MORPH-II that we address here.…”

Section: Introductionmentioning

confidence: 99%

Preliminary Studies on a Large Face Database

Yip

Bingham

Kempfert

et al. 2018

2018 IEEE International Conference on Big Data (Big Data)

View full text Add to dashboard Cite

We perform preliminary studies on a large longitudinal face database MORPH-II, which is a benchmark dataset in the field of computer vision and pattern recognition. First, we summarize the inconsistencies in the dataset and introduce the steps and strategy taken for cleaning. The potential implications of these inconsistencies on prior research are introduced. Next, we propose a new automatic subsetting scheme for evaluation protocol. It is intended to overcome the unbalanced racial and gender distributions of MORPH-II, while ensuring independence between training and testing sets. Finally, we contribute a novel global framework for age estimation that utilizes posterior probabilities from the race classification step to compute a racecomposite age estimate. Preliminary experimental results on MORPH-II are presented.

show abstract

Giant panda age recognition based on a facial image deep learning system

Hou

et al. 2022

Ecology and Evolution

View full text Add to dashboard Cite

The conservation of the giant panda ( Ailuropoda melanoleuca ), as an iconic vulnerable species, has received great attention in the past few decades. As an important part of the giant panda population survey, the age distribution of giant pandas can not only provide useful instruction but also verify the effectiveness of conservation measures. The current methods for determining the age groups of giant pandas are mainly based on the size and length of giant panda feces and the bite value of intact bamboo in the feces, or in the case of a skeleton, through the wear of molars and the growth line of teeth. These methods have certain flaws that limit their applications. In this study, we developed a deep learning method to study age group classification based on facial images of captive giant pandas and achieved an accuracy of 85.99% on EfficientNet. The experimental results show that the faces of giant pandas contain some age information, which mainly concentrated between the eyes of giant pandas. In addition, the results also indicate that it is feasible to identify the age groups of giant pandas through the analysis of facial images.

show abstract

Effective training of convolutional neural networks for face-based gender and age prediction

Cited by 121 publications

References 20 publications

Transfer Learning with Deep CNNs for Gender Recognition and Age Estimation

Transfer Learning with Deep CNNs for Gender Recognition and Age Estimation

Preliminary Studies on a Large Face Database

Giant panda age recognition based on a facial image deep learning system

Contact Info

Product

Resources

About