Age Group and Gender Estimation in the Wild with Deep RoR Architecture

Automatically predicting age group and gender from face images acquired inunconstrained conditions is an important and challenging task in manyreal-world applications. Nevertheless, the conventional methods withmanually-designed features on in-the-wild benchmarks are unsatisfactory becauseof incompetency to tackle large variations in unconstrained images. Thisdifficulty is alleviated to some degree through Convolutional Neural Networks(CNN) for its powerful feature representation. In this paper, we propose a newCNN based method for age group and gender estimation leveraging ResidualNetworks of Residual Networks (RoR), which exhibits better optimization abilityfor age group and gender classification than other CNN architectures.Moreover,two modest mechanisms based on observation of the characteristics of age groupare presented to further improve the performance of age estimation.In order tofurther improve the performance and alleviate over-fitting problem, RoR modelis pre-trained on ImageNet firstly, and then it is fune-tuned on theIMDB-WIKI-101 data set for further learning the features of face images,finally, it is used to fine-tune on Adience data set. Our experimentsillustrate the effectiveness of RoR method for age and gender estimation in thewild, where it achieves better performance than other CNN methods. Finally, theRoR-152+IMDB-WIKI-101 with two mechanisms achieves new state-of-the-art resultson Adience benchmark.