Comprehensive and accurate eye modeling is crucial to a variety of applications, including human-computer interaction, assistive technologies, and medical diagnosis. However, most studies focus on the localization of one or two components of eyes, such as pupil or iris, lacking a comprehensive eye model. We propose to model an eye image by a set of parametric curves. The set of curves are plotted on an eye image to form a Contour-Eye image. A deep neural network is trained to evaluate the fitness of the Contour-Eye image. Then an evolutionary process is conducted to search the best fitting curve set, guided by the trained deep neural network. Finally, an accurate eye model with optimized parametric curves is obtained. For the algorithm evaluation, a finely annotated eye dataset denoted as FAED-50 is established by us, which contains 2,498 eye images from 50 subjects. The experimental results on the FAED-50 and the relabeled CASIA datasets and comparison with the state-of-the-art methods demonstrate the effectiveness and accuracy of the proposed parametric model. Copyright © 2020 Published by Elsevier Ltd.
CitationZheng, Y., Fu, H., Li, R., Hsung, T.-C., Song, Z., & Wen, D. (2021). Deep neural network oriented evolutionary parametric eye modeling. Pattern Recognition, 113. Retrieved from https://doi.org/10.1016/j.patcog.2020.107755
- Parametric eye modeling
- Deep neural networks
- Evolutionary search
- Fitness evaluation