A Deep Learning Approach to Recognize Faces After Plastic Surgery

Facial recognition is a challenging research area since the past decade due to exponentially increasing intervention of humans to achieve anything and everything using technology. Facial recognition is a task; that is, trivially accomplished by human beings, even under fluctuating light and when faces are transformed by age or hindered by accessories or facial hair. Nevertheless, it has remained a perplexing computer vision challenge for decades. Due to upsurge in acceptance and accomplishment of deep learning models in computer vision, features extracted by convolutional neural networks (CNNs) can be employed for facial recognition. However, all the techniques involving deep learning require huge amounts of data, thereby increasing the computational requirements of any algorithm. In this paper, a method has been introduced with minimal computational requirements and minimal time complexity for identifying and recognizing faces which are surgically altered. Recognition is done by introducing a simple CNN for facial trait extraction which is followed by performing the classification task. Plastic surgery facial dataset is used for training, testing and validating the proposed deep network. This indeed is very beneficial for ensuring safety and security of an individual’s identity as it would assist to recognize offenders, impersonators or anyone who conceals their individuality.