Age and Gender Recognition on Imbalanced Dataset of Face Images with Deep Learning

The paper describes usage of deep neural networks based on ResNet and Xception architectures for recognition of age and gender of imbalanced dataset of face images. Described dataset collection process from open sources. Training sample contains more than 210000 images. Testing sample have more 1700 special selected face images with different ages and genders. Training data has imbalanced number of images per class. Accuracy for gender classification and mean absolute error for age estimation are used to analyze results quality. Age recognition is described as classification task with 101 classes. Gender recognition is solved as classification task with two categories. Paper contains analysis of different approaches to data balancing and their influence to recognition results. The computing experiment was carried out on a graphics processor using NVidia CUDA technology. The average recognition time per image is estimated for different deep neural networks. Obtained results can be used in software for public space monitoring, collection of visiting statistics etc.