Deep Real-world and Real-time Face Identification System

Face identification has been one of the most challenging and attractive areas, and it has become a popular research area in the computer vision community because of its impact in application areas, such as surveillance, security systems, and biometrics. With the introduction deep learning technique, face recognition accuracy has skyrocketed. The state-of-the-art has achieved close to 100% recognition rates by applying convolutional neural networks (CNN), as these networks are more robust in regards to image variation than other methods are. However, in an uncontrolled environment with illumination, extreme poses, and variations in facial expression, the face recognition problem is far from being solved. This paper has two contributions, the first one is to discuss real-world challenges in face identification and suggest solutions for them. To overcome one these challenges, we train an MLP model to validate the output of CNN which reduce the false positive rate. The second contribution is to implement an OpenVX computational graph model to achieve real-time performance and reduce the run-time.