Soft Computing-Based Approach for Face Recognition on Plastic Surgery and Skin Colour-Based Scenarios Using CNN

Abstract

AbstractIn recent years, face recognition has become one of the most common technologies in the identification of humans which is widely used as biometrics. With many of the biometric systems that are already out there, facial recognition has become one of the most important technologies for quickly identifying people without having to ask them. This will not cause any unnecessary delays. However, there are several face-recognition systems that rely on traditional machine learning as well. But they do not work when it comes to things like facial expressions, posture, occlusion, and scale. The face-recognition method described in this article employs a convolutional neural network (CNN) to locate faces within a picture. Viola–Jones face detection is used to locate faces in a picture, and a pre-trained CNN extracts facial traits from the faces automatically. However, a large database of facial images is made in order to have more images for each subject and to include different noise levels for the best training of CNNs. Overall accuracy of the system was increased to 98.81% which will depict the effectiveness of deep face recognition. We attempt to design an algorithm that overcomes the difficulties associated with facial identification owing to skin by feeding faces into a CNN for feature extraction and then passing them to a softmax classifier for classification. Also to develop an algorithm to overcome the challenges that occur in the face recognition due to external and artificial features like plastic surgery.KeywordsCNNFace recognitionArtificial featuresMachine learning