Simultaneous Facial Age Group and Gender Recognition using Efficient Local-Global Attention Network for Intelligent Advertising
Adri Priadana, Duy-Linh Nguyen, Xuan-Thuy Vo, Muhamad Dwisnanto Putro, G. F. Cao, Kang-Hyun Jo · 2024
Facial age group and gender recognition have attracted significant interest due to their wide range of applications and practical uses, including aiding advertising platforms in delivering relevant content. Performing efficient simultaneous recognition of facial age group and gender is crucial for these applications, necessitating seamless performance on inexpensive devices to mitigate implementation expenses. This work introduces an Efficient Local-Global Attention Network (ELGA-Net) for facial age group and gender recognition simultaneously. It proposes an Efficient Local-Global Attention (ELGA) block to enhance the quality of feature maps locally while capturing global contextual information by learning relationships between different parts of the feature maps. As a result, the proposed network achieves competitive performance on the UTKFace dataset. Moreover, it attains real-time speed at 113 frames per second (FPS) on an Intel Core i7-9750H CPU 2.6 GHz device when integrated with face detection as an initial process.