Facial Based Gender Classification for Real Time Applications

Appearance and facial features play an important role in gender recognition through images. For gender classification, multiple techniques were presented to acquire better results in which preprocessing part is one of the major and very important for gender classification as it removes noise, enhances, images, and eliminates any unnatural colors from an image. Another major aspect is the efficient feature extraction method. If features extracted accurately then the result of classification will improve. Over the past few years, gender classification techniques work perfectly for a controlled environment. However, challenges occurred for real-time applications due to low resolution, off-angle poses, faces with occlusion, and various expressions. The main focus of this study is to overcome existing challenges and propose a method that can be implemented in real-time applications. This research work proposed a novel method in which CNN has been used for classification of gender for real-time application. To assess the performance of proposed method experiments were conducted on static images and video data sets.  The proposed research work achieved 98% of accuracy during the experiments.