Acta Optica Sinica, Volume. 40, Issue 11, 1111002(2020)
Infrared Target Modeling Method Based on Double Adversarial Autoencoding Network
Figures & Tables(10)
Fig. 6. Test results of CDAAE model on infrared dataset. (a) Real samples; (b) reconstructed samples; (c) random samples
Fig. 7. Test results of CDAAE model on MINIST dataset. (a) Real samples; (b) reconstructed samples; (c) random samples
Table 1. Hyper parameters used for CDAAE model training
View tableTable 1. Hyper parameters used for CDAAE model training
Hyper parameter Infrared dataset MINIST dataset Training loop 100000 100000 Class number 4 10 Batch size 64 256 Slope of Leaky ReLU 0.01 0.01 Learning rate of E/Dg 0.00002 0.00002 Learning rate of Dl/Ds 0.0001 0.0001 Data dimension in latent-space 8 4 Optimization of E/Dg/Dl/Ds per training loop 2∶4∶1∶1 2∶4∶1∶1
Table 2. FID evaluation results of CDAAE model
View tableTable 2. FID evaluation results of CDAAE model
Condition Infrared dataset MINIST dataset Group 1 vs group 2 10.13 8.85 Real samples versus reconstructed samples 14.26 7.51 Real samples versus random samples 35.36 10.97 Real samples versus random samples from α-GAN 32.47 12.18 Real samples versus random samples from C-DCGAN 48.83 27.62 Real samples versus random samples from CVAE 52.74 25.55
Table 3. Accuracy in data augmentation process%
View tableTable 3. Accuracy in data augmentation process%
Number of pictures in Infrared dataset Number of pictures in MINIST dataset Infrared dataset MINIST dataset 1200 3000 92.83 97.80 1200+1200 3000+3000 94.33 98.33 1200+2400 3000+6000 96.00 99.26
Tools
Get Citation
Copy Citation Text
Zhuang Miao, Yong Zhang, Weihua Li. Infrared Target Modeling Method Based on Double Adversarial Autoencoding Network[J]. Acta Optica Sinica, 2020, 40(11): 1111002
Paper Information
Category: Imaging Systems
Received: Jan. 9, 2020
Accepted: Mar. 10, 2020
Published Online: Jun. 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