Aprile 5, 2014

Lost password

Dataset

These videos and images have been used in order to test our methods for gender recognition in different scientific works.

UNISA-Public dataset is composed by 12 video sequences acquired in the University of Salerno in real environments. It was collected with the purpose of evaluating gender recognition algorithms in real life scenarios, since the images extracted from videos are surely more challenging than the ones provided by the classic datasets. We also provide the face images extracted with OpenCV and Matlab.

GENDER-FERET dataset is a balanced subset of the FERET dataset, adapted for gender recogntion purposes. It consists of 946 grayscale images, already divided in training set (237 m, 237 f) and test set (236 m, 236 f).

GENDER-COLOR-FERET dataset is a balanced subset of the COLOR-FERET dataset, adapted for gender recogntion purposes. In this case the images are coloured and the dataset is composed by 836 faces. The dataset is completely balanced, since both the training and the test set are composed of 209 male and 209 female faces.


References

Papers on the topic:

  • A Convolutional Neural Network for Gender Recognition Optimizing the Accuracy/Speed Tradeoff Gender recognition has been among the most investigated problems in the last years; although several contributions have been proposed, gender recognition in unconstrained environments is still a challenging problem and a definitive solution has not been found yet. Furthermore, Deep Convolutional Neural Networks (DCNNs) achieve very interesting performance, but they typically require a huge amount of computational resources (CPU, GPU, RAM, storage), that are not always available in real systems, due to their cost or to specific application constraints (when the application needs to be installed directly on board of low-power smart cameras, e.g. for digital signage). In the latest years the Machine Learning community developed an interest towards optimizing the efficiency of Deep Learning solutions, in order to make them portable and widespread. In this work we propose a compact DCNN architecture for Gender Recognition from face images that achieves approximately state of the art accuracy at a highly reduced computational cost (almost five times). We also perform a sensitivity analysis in order to show how some changes in the architecture of the network can influence the tradeoff between accuracy and speed. In addition, we compare our optimized architecture with popular efficient CNNs on various common benchmark dataset, widely adopted in the scientific community, namely LFW, MIVIA-Gender, IMDB-WIKI and Adience, demonstrating the effectiveness of the proposed solution.
  • An effective real time gender recognition system for smart cameras In recent years we have assisted to a growing interest for embedded vision, due to the availability of low cost hardware systems, effective for energy consumption, flexible for their size at the cost of limited (compared to the server) computing resources. Their use is boosted by the simplicity of their positioning in places where energy or network bandwidth is limited. Smart cameras are digital cameras embedding computer systems able to host video applications; due to the cost and the performance, they are progressively gaining popularity and conquering large amount of the market. Smart cameras are now able to host on board video applications, even if this imposes an heavy reformulation of the algorithms and of the software design so as to make them compliant with the limited CPUs and the small RAM and flash memory (typically of a few megabytes). In this paper we propose a method for gender recognition on video sequences, specifically designed for making it suited to smart cameras; although the algorithm uses very limited resources (in terms of RAM and CPU), it is able to run on smart cameras available today, presenting at the same time an high accuracy on unrestricted videos taken in real environments (malls, shops, etc.).
  • Fusion of domain-specific and trainable features for gender recognition from face images. The popularity and the appeal of systems which are able to automatically determine the gender from face images is growing rapidly. Such a great interest arises from the wide variety of applications, especially in the fields of retail and video surveillance. In recent years there have been several attempts to address this challenge, but a definitive solution has not yet been found. In this paper we propose a novel approach that fuses domain-specific and trainable features to recognize the gender from face images. In particular, we use the SURF descriptors extracted from 51 facial landmarks related to eyes, nose and mouth as domain dependent features, and the COSFIRE filters as trainable features. The proposed approach turns out to be very robust with respect to the well known face variations, including different poses, expressions and illumination conditions. It achieves state-of-the-art recognition rates on the GENDER-FERET (94.7%) and on the LFW (99.4%) datasets, which are two of the most popular benchmarks for gender recognition. We further evaluated the method on a new dataset acquired in real scenarios, the UNISA-Public, recently made publicly available. It consists of 206 training (144 male, 62 female) and 200 test (139 male, 61 female) images that are acquired with a real-time indoor camera capturing people in regular walking motion. Such experiment has the aim to assess the capability of the algorithm to deal with face images extracted from videos, which are definitely more challenging than the still images available in the standard datasets. Also for this dataset we achieved a high recognition rate of 91.5%, that confirms the generalization capabilities of the proposed approach. Of the two types of features, the trainable COSFIRE filters are the most effective and, given their trainable character, they can be applied in any visual pattern recognition problem.
  • Digital Signage by Real-Time Gender Recognition From Face Images. Digital signage is a new advertising strategy using smart multimedia screens, which carries out the dynamic customization of the promotional content according to the customers who are looking at the monitor. Gender recognition from face images is among the most popular applications for digital signage, since it allows to select in real-time advertising spots customized for males or females. In this paper, we propose a system which implements this solution using a smart camera mounted above the monitor, dedicated to gender recognition in real-time, and a component that dynamically modifies the content projected on the screen according to the gender of the audience. The computer vision algorithm is designed to be as fast as effective, since the whole processing chain must be performed in real-time in order to avoid missing people passing in front of the screen. We evaluated the performance of the proposed solution on a standard dataset for gender recognition in the wild and in a real fair, obtaining a gender recognition accuracy of 94.99% and 92.70%, respectively, that is very relevant in such unconstrained scenarios. In addition, the method is able to process 5 fps on a smart camera and, thus, it can be used in a digital signage application.
  • A system for gender recognition on mobile robots The combination of artificial intelligence and robotics opens the way to disruptive future developments in the industrial and collaborative robotics. The recent advances of the deep learning technologies materialize the possibility to provide a robot perceptive and reasoning skills and, consequently, the capability to autonomously interact with a human. In this paper we ride the wave of intelligent robotics by designing an autonomous robot able to recognize the gender of the customers in a shopping center and to interact with them proposing customized advertising and promotional material. We train two well-known Convolutional Neural Network architectures to recognize gender from face images. In order to run them in real time we extend the computational capabilities of a social robotics platform with an embedded parallel computation accelerator. The experimental analysis, carried out on video sequences acquired in real scenarios, demonstrate the suitability of the proposed platform for the considered social robotics application in terms of both latency and accuracy.
  • Gender recognition from face images using trainable shape and color features Gender recognition from face images is an important application and it is still an open computer vision problem, even though it is something trivial from the human visual system. Variations in pose, lighting, and expression are few of the problems that make such an application challenging for a computer system. Neurophysiological studies demonstrate that the human brain is able to distinguish men and women also in absence of external cues, by analyzing the shape of specific parts of the face. In this paper, we describe an automatic procedure that combines trainable shape and color features for gender classification. In particular the proposed method fuses edge-based and color-blob-based features by means of trainable COSFIRE filters. The former types of feature are able to extract information about the shape of a face whereas the latter extract information about shades of colors in different parts of the face. We use these two sets of features to create a stacked classification SVM model and demonstrate its effectiveness on the GENDER-COLOR-FERET dataset, where we achieve an accuracy of 96.4%.
  • Fast gender recognition in videos using a novel descriptor based on the gradient magnitudes of facial landmarks. The growing interest in recent years for gender recognition from face images is mainly attributable to the wide range of possible applications that can be used for commercial and marketing purposes. It is desirable that such algorithms process high resolution video frames acquired by using surveillance cameras in real-time. To the best of our knowledge, however, there are no studies which analyze the computational impact of the methods and the difficulties related to the processing of faces extracted from videos captured in the wild. We propose a novel face descriptor based on the gradient magnitudes of facial landmarks, which are points automatically extracted from the face contour, eyes, eyebrows, nose, mouth and chin. We evaluate the effectiveness and efficiency of the proposed approach on two new datasets, which we made available online and that consist of color face images and color video sequences acquired in real scenarios. The proposed approach is more efficient and effective than three commercial libraries.
  • Gender recognition from face images using a fusion of svm classifiers. The recognition of gender from face images is an important application, especially in the fields of security, marketing and intelligent user interfaces. We propose an approach to gender recognition from faces by fusing the decisions of SVM classifiers. Each classifier is trained with different types of features, namely HOG (shape), LBP (texture) and raw pixel values. For the latter features we use an SVM with a linear kernel and for the two former ones we use SVMs with histogram intersection kernels. We come to a decision by fusing the three classifiers with a majority vote. We demonstrate the effectiveness of our approach on a new dataset that we extract from FERET. We achieve an accuracy of 92.6 %, which outperforms the commercial products Face++ and Luxand.
  • Gender recognition from face images with trainable COSFIRE filters. Gender recognition from face images is an important application in the fields of security, retail advertising and marketing. We propose a novel descriptor based on COSFIRE filters for gender recognition. A COSFIRE filter is trainable, in that its selectivity is determined in an automatic configuration process that analyses a given prototype pattern of interest. We demonstrate the effectiveness of the proposed approach on a new dataset called GENDER-FERET with 474 training and 472 test samples and achieve an accuracy rate of 93.7%. It also outperforms an approach that relies on handcrafted features and an ensemble of classifiers. Furthermore, we perform another experiment by using the images of the Labeled Faces in the Wild (LFW) dataset to train our classifier and the test images of the GENDER-FERET dataset for evaluation. This experiment demonstrates the generalization ability of the proposed approach and it also outperforms two commercial libraries, namely Face++ and Luxand.

Download

In order to download the datasets click here.


Help

If you have any problems, do not hesitate to contact us here.