Real-time embedded age and gender classification in unconstrained video

Abstract

Recently, automatic demographic classification has found its way into embedded applications such as targeted advertising in mobile devices, and in-car warning systems for elderly drivers. In this thesis, we present a complete framework for video-based gender classification and age estimation which can perform accurately on embedded systems in real-time and under unconstrained conditions. We propose a segmental dimensionality reduction technique utilizing Enhanced Discriminant Analysis (EDA) to minimize the memory and computational requirements, and enable the implementation of these classifiers for resource-limited embedded systems which otherwise is not achievable using existing resource-intensive approaches. On a multi-resolution feature vector we have achieved up to 99.5% compression ratio for training data storage, and a maximum performance of 20 frames per second on an embedded Android platform.Also, we introduce several novel improvements such as face alignment using the nose, and an illumination normalization method for unconstrained environments using bilateral filtering. These improvements could help to suppress the textural noise, normalize the skin color, and rectify the face localization errors. A non-linear Support Vector Machine (SVM) classifier along with a discriminative demography-based classification strategy is exploited to improve both accuracy and performance of classification. We have performed several cross-database evaluations on different controlled and uncontrolled databases to assess the generalization capability of the classifiers. Our experiments demonstrated competitive accuracies compared to the resource-demanding state-of-the-art approaches.ii Acknowledgements I would like to take the opportunity to appreciate my supervisors, Prof. Robert Laganière, for his invaluable guidance, encouragement, and persistent support throughout the project and, Prof. WonSook Lee, for providing me with all the necessary facilities for research, and sharing expertise.

A Gender Recognition System from Facial Image

“…Extracting features using tCENTRIST shows 93.75% accuracy and with BHEP it increases to 94.29% in LFW dataset which is better than [32]. Again in Adience dataset proposed method shows better accuracy than [3] in both cases. Our method found 97.24% and 98.30% accuracies in Color FERET dataset without using BHEP and with BHEP respectively.…”

“…Although they proposed these two methods for garments texture classification, these two also can be used in different facial image analysis research. Azarmehr et al [3] applied Multi-scale Local Binary Patterns (MSLBP) proposed by [18] for gender recognition. They proposed to apply a Bilateral filtering approach in preprocessing to suppress the noises.…”

An efficient approach for textual data classification using deep learning

“…This work on smoothing approaches uses a multinomial model for brief texts. This linear approach has recently been discovered to be capable of overcoming the dimensionality curse and delivering real-time performance (Azarmehr et al, 2015 ).…”

REGA: Real-Time Emotion, Gender, Age Detection Using CNN—A Review

“…Ramin Azarmehr , Robert Laganiere , Won-Sook Lee et al [9] proposed a system using EDA achieved 99% and for better accuracy and performance use support vector machine (SVM) and demographic classification strategies.…”