Two-tiered face verification with low-memory footprint for mobile devices

Mobile devices had their popularity and affordability greatly increased in recent years. As a consequence of their ubiquity, these devices now carry all sorts of personal data that should be accessed only by their owner. Even though knowledge-based procedures are still the main methods to secure the owner’s identity, recently biometric traits have been employed for a more secure and effortless authentication. In this work, we propose a facial verification method optimized to the mobile environment. It consists of a two-tiered procedure that combines hand-crafted features and a convolutional neural network (CNN) to verify if the person depicted in a photo corresponds to the device owner. To train a CNN for the verification task, we propose a hybrid-image input, which allows the network to process encoded information of a pair of face images. Our experiments show that the proposed solution outperforms state-of-the-art face verification methods, providing a 4x speedup when processing an image in recent smartphone models. Additionally, we show that the two-tiered procedure can be coupled with existing face verification CNNs improving their accuracy and efficiency. We also present a new dataset of selfie pictures – RCD dataset – that hopefully will support future research in this scenario.