2021 IEEE International Conference on Acoustics, Speech and Signal Processing

6-11 June 2021 • Toronto, Ontario, Canada

Extracting Knowledge from Information

2021 IEEE International Conference on Acoustics, Speech and Signal Processing

6-11 June 2021 • Toronto, Ontario, Canada

Extracting Knowledge from Information

Technical Program

Paper Detail

Paper IDIVMSP-18.3
Paper Title CONTINUOUS FACE AGING GENERATIVE ADVERSARIAL NETWORKS
Authors Seogkyu Jeon, Pilhyeon Lee, Kibeom Hong, Hyeran Byun, Yonsei University, South Korea
SessionIVMSP-18: Faces in Images & Videos
LocationGather.Town
Session Time:Wednesday, 09 June, 16:30 - 17:15
Presentation Time:Wednesday, 09 June, 16:30 - 17:15
Presentation Poster
Topic Image, Video, and Multidimensional Signal Processing: [IVARS] Image & Video Analysis, Synthesis, and Retrieval
IEEE Xplore Open Preview  Click here to view in IEEE Xplore
Virtual Presentation  Click here to watch in the Virtual Conference
Abstract Face aging is the task aiming to translate the faces in input images to designated ages. For problem simplification, previous methods build age groups, each of which consists of ten years. This limits them only able to produce discrete results for the pre-defined target age groups. Therefore, the exact ages of the translated results are unknown and it is unable to obtain the faces of intermediate ages between groups. To overcome these limitations, we propose the continuous face aging generative adversarial networks (CFA-GAN). In specific, to make the continuous aging feasible, we propose to decompose image features into two orthogonal features: the identity basis feature and the age basis feature. They are in turn trained to contain appropriate information and allow for handling the continuous age attributes. Moreover, we introduce the novel loss function for identity preservation which maximizes the cosine similarity between the original and the generated identity basis features. With the qualitative and quantitative evaluations on the benchmark face dataset, we demonstrate the smooth and continuous aging ability of our model, showing its effectiveness. To the best of our knowledge, this work is the first attempt to handle continuous target ages, instead of discrete age groups.