Multi-scale feature fusion model followed by residual network for generation of face aging and de-aging

Face aging is one of the most interesting style transfer ideas due to the extraordinary development in image synthesis succeeded by deep learning models that is the generative adversarial networks and its marvelous impact on practical applications such as finding missing child after few years, smart voting where we have to update the data based on the age changes of people. The existing face aging methods have proven the achievement in the case of the paired image dataset. Collecting the paired data samples of different age groups is hard and expensive. Encouraged by GAN's success in a variety of fields for image-to-image conversion problems. The main aim of this paper is to keep the original identity as it is in the face aging problem. We have designed an approach known as the multi-scale feature fusion model followed by a residual network to generate images of a person based on different age conditions. We worked on an unpaired image dataset because we do not have the dataset of the same person in different age categories. In this paper, we have used the UTKFace dataset which is publicly available. The scope of research is to consider only two age categories. The results are obtained by performing experiments and through a survey of people which indicates the modern method for face age progression and regression.