Acta Optica Sinica, Volume. 40, Issue 5, 0510001(2020)
Facial Image Translation in Short-Wavelength Infrared and Visible Light Based on Generative Adversarial Network
Figures & Tables(9)
Fig. 2. Structure of generator and discriminator. (a) Generator; (b) discriminator
Fig. 4. Comparison of generated images by different frameworks. (a) Short-wave images; (b) proposed framework; (c) DRIT; (d) UNIT; (e) MUNIT; (f) CycleGAN; (g) DualGAN; (h) reference images
Table 1. Results of identity loss function on Facenet
View tableTable 1. Results of identity loss function on Facenet
Model FID Accuracy (all) Accuracy (frontal) CycleGAN 48.2173 0.137 0.146 CycleGAN+Lidentity 65.0052 0.220 0.267
Table 2. Index analysis of overall loss function on Facenet
View tableTable 2. Index analysis of overall loss function on Facenet
α2 FID Accuracy (all) Accuracy (frontal) 0 1.0 5.0 10.0 12.5 15.0 20.0 50.0 48.2173 52.9253 41.6498 41.3721 43.4405 43.1670 38.4180 65.9726 0.137 0.121 0.133 0.152 0.144 0.161 0.142 0.145 0.146 0.119 0.146 0.179 0.165 0.175 0.167 0.154
Table 3. Index analysis of synthesis-reconstruction loss function on Facenet
View tableTable 3. Index analysis of synthesis-reconstruction loss function on Facenet
α3 FID Accuracy (all) Accuracy (frontal) 0 0.05 0.50 1.00 5.00 50.00 48.2173 41.4332 52.5602 58.1832 87.7636 105.1097 0.137 0.168 0.252 0.234 0.216 0.034 0.146 0.214 0.311 0.294 0.243 0.029
Table 4. Results of proposed framework with different parameters
View tableTable 4. Results of proposed framework with different parameters
α2 α3 FID Accuracy (all) Accuracy (frontal) 0 10 15 8 10 10 11 15 0 0.05 0.05 0.50 0.40 0.50 0.50 0.50 48.2173 43.7915 39.2064 48.6155 43.9396 47.3351 49.9668 48.2815 0.137 0.234 0.200 0.297 0.260 0.334 0.290 0.271 0.146 0.288 0.216 0.370 0.325 0.409 0.393 0.354
Table 5. Index analysis of results generated by different frameworks
View tableTable 5. Index analysis of results generated by different frameworks
Framework FID Accuracy (all) Accuracy (frontal) Real image Proposed framework CycleGAN DualGAN UNIT MUNIT DRIT 16.7590 47.3351 48.2173 46.7556 50.9268 69.1422 52.6906 0.832 0.334 0.137 0.078 0.166 0.112 0.222 0.409 0.146 0.086 0.214 0.129 0.290
Tools
Get Citation
Copy Citation Text
Linmiao Hu, Yong Zhang. Facial Image Translation in Short-Wavelength Infrared and Visible Light Based on Generative Adversarial Network[J]. Acta Optica Sinica, 2020, 40(5): 0510001
Paper Information
Category: Image Processing
Received: Sep. 23, 2019
Accepted: Nov. 9, 2019
Published Online: Mar. 10, 2020
The Author Email: Zhang Yong (zybxy@sina.com)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20