Bidirectional Face Aging Synthesis Based on Improved Deep Convolutional Generative Adversarial Networks
AbstractThe use of computers to simulate facial aging or rejuvenation has long been a hot research topic in the field of computer vision, and this technology can be applied in many fields, such as customs security, public places, and business entertainment. With the rapid increase in computing speeds, complex neural network algorithms can be implemented in an acceptable amount of time. In this paper, an optimized face-aging method based on a Deep Convolutional Generative Adversarial Network (DCGAN) is proposed. In this method, an original face image is initially mapped to a personal latent vector by an encoder, and then the personal potential vector is combined with the age condition vector and the gender condition vector through a connector. The output of the connector is the input of the generator. A stable and photo-realistic facial image is then generated by maintaining personalized facial features and changing age conditions. With regard to the objective function, the single adversarial loss of the Generated Adversarial Network (GAN) with the perceptual similarity loss is replaced by the perceptual similarity loss function, which is the weighted sum of adversarial loss, feature space loss, pixel space loss, and age loss. The experimental results show that the proposed method can synthesize an aging face with rich texture and visual reality and outperform similar work. View Full-Text
Share & Cite This Article
Liu, X.; Zou, Y.; Xie, C.; Kuang, H.; Ma, X. Bidirectional Face Aging Synthesis Based on Improved Deep Convolutional Generative Adversarial Networks. Information 2019, 10, 69.
Liu X, Zou Y, Xie C, Kuang H, Ma X. Bidirectional Face Aging Synthesis Based on Improved Deep Convolutional Generative Adversarial Networks. Information. 2019; 10(2):69.Chicago/Turabian Style
Liu, Xinhua; Zou, Yao; Xie, Chengjuan; Kuang, Hailan; Ma, Xiaolin. 2019. "Bidirectional Face Aging Synthesis Based on Improved Deep Convolutional Generative Adversarial Networks." Information 10, no. 2: 69.
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.