Descriptors and regions of interest fusion for in- and cross-database gender classification in the wild

Abstract

Gender classification (GC) has achieved high accuracy in different experimental evaluations based mostly on inner facial details. However, these results do not generalize well in unrestricted datasets and particularly in cross-database experiments, where the performance drops drastically. In this paper, we analyze the state-of-the-art GC accuracy on three large datasets: MORPH, LFW and GROUPS. We discuss their respective difficulties and bias, concluding that the most challenging and wildest complexity is present in GROUPS. This dataset covers hard conditions such as low resolution imagery and cluttered background. Firstly, we analyze in depth the performance of different descriptors extracted from the face and its local context on this dataset. Selecting the bests and studying their most suitable combination allows us to design a solution that beats any previously published results for GROUPS with the Dago's protocol, reaching an accuracy over 94.2%, reducing the gap with other simpler datasets. The chosen solution based on local descriptors is later evaluated in a cross-database scenario with the three mentioned datasets, and full dataset 5-fold cross validation. The achieved results are compared with a Convolutional Neural Network approach, achieving rather similar marks. Finally, a solution is proposed combining both focuses, exhibiting great complementarity, boosting GC performance to beat previously published results in GC both cross-database, and full in-database evaluations.

Multi-scale score level fusion of local descriptors for gender classification in the wild

“…The reader must observe that the evaluation protocol is not exactly the same in most works. Indeed, it is not [8] 89.8 [9] 91.6 [36] 90.6 [19] 91.6 [11] 92.4 [31] Detected faces 2 86.4 [13] 90.4 [6] Adults 3 80.5 [23] Full dataset 87.1 [9] 90.8 [42] LFW Subset 4 94.8 [45] 98.0 [36] BEFIT 5 96.2 [15] 94.0 [40] Half dataset 6 98.0 [6] Full dataset 79.5 [41] 94.6 [28] 96.9…”

“…Our baseline is given by our recent results that compete, as far as we know, the state of the art in facial based GC in the wild [8,9,11], enclosing a comparison with CNN approaches.…”

Gender Privacy: An Ensemble of Semi Adversarial Networks for Confounding Arbitrary Gender Classifiers

“…Extracting soft biometric attributes, such as age and gender from face images, has been extensively studied [9,16,13]. A wide range of methods have been employed, including those based on custom feature extraction techniques [10] and those based on deep learning techniques [26,29,20,12,16]. However, imparting soft biometric privacy by confounding such attributes is a relatively recent research area.…”

Transfer Learning with Deep CNNs for Gender Recognition and Age Estimation

“…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.…”