Acta Optica Sinica, Volume. 40, Issue 18, 1810001(2020)
Infrared Simulation Based on Cascade Multi-Scale Information Fusion Adversarial Network
Figures & Tables(12)
Fig. 5. Infrared simulation images generated by different algorithms. (a) Car; (b) bicycle rider; (c) building; (d) long-distance vehicle
Table 1. Detailed configuration about G1 and Gs
View tableTable 1. Detailed configuration about G1 and Gs
Encoder (G1/Gs filters) Number of channels G1/Gs Decoder (G1/ filters) Number of channels G1/Gs dconv16 (3/3) 3/3 conv1 (64/4) 64/4 dconv15 (64/4) 128/8 conv2 (128/8) 128/8 dconv14 (128/8) 256/16 conv3 (256/16) 256/16 dconv13 (256/16) 512/32 conv4 (512/32) 512/32 dconv12 (512/32) 1024/64 conv5 (512/32) 512/32 dconv11 (512/32) 1024/64 conv6 (512/32) 512/32 dconv10 (512/32) 1024/64 conv7 (512/32) 512/32 dconv9 (512/32) 1024/64 conv8 (512/32) 512/32
Table 2. Detailed configuration about discriminator
View tableTable 2. Detailed configuration about discriminator
Network Input/output channels Stride BN Activation function conv1 6/64 2 N L conv2 64128 2 Y L conv3 128/256 2 Y L conv4 256/512 2 Y L conv5 512/512 1 Y L conv6 512/1 1 N S
Table 3. Comparison of objective indicators of algorithms
View tableTable 3. Comparison of objective indicators of algorithms
Method The lower, the better The higher, the better Rel Avg log10 RMS δ<1.25 PSNR SSIM FCRN[ 11 ]0.286 0.144 1.060 0.409 21.204 0.962 FLED-Net[ 12 ]0.238 0.100 0.853 0.602 22.921 0.987 Pix2pix[ 20 ]0.248 0.107 0.906 0.571 22.431 0.985 Selection-GAN[ 23 ]0.284 0.112 0.958 0.554 21.976 0.982 Proposed 0.257 0.102 0.876 0.612 22.657 0.989
Table 4. Comparison of first level network and CMIF-GAN
View tableTable 4. Comparison of first level network and CMIF-GAN
Method The lower, the better The higher, the better Rel Avg log10 RMS δ<1.25 PSNR SSIM First level 0.265 0.107 0.918 0.589 22.310 0.987 Proposed 0.257 0.102 0.876 0.612 22.657 0.989
Table 5. Comparison of auxiliary tasks experiments
View tableTable 5. Comparison of auxiliary tasks experiments
Setup The lower, the better The higher, the better Rel Avg log10 RMS δ<1.25 PSNR SSIM -Gs,Gr 0.268 0.107 0.912 0.586 22.321 0.988 -Gg 0.276 0.106 0.925 0.594 22.280 0.986 -Gs 0.264 0.105 0.901 0.598 22.510 0.987 Proposed 0.257 0.102 0.876 0.612 22.657 0.989
Table 6. Comparison of MFM module experiments
View tableTable 6. Comparison of MFM module experiments
Setup The lower, the better The higher, the better Rel Avg log10 RMS δ<1.25 PSNR SSIM -MFM 0.268 0.105 0.905 0.600 22.465 0.987 Proposed 0.257 0.102 0.876 0.612 22.657 0.989
Tools
Get Citation
Copy Citation Text
Ruiming Jia, Tong Li, Shengjie Liu, Jiali Cui, Fei Yuan. Infrared Simulation Based on Cascade Multi-Scale Information Fusion Adversarial Network[J]. Acta Optica Sinica, 2020, 40(18): 1810001
Paper Information
Category: Image Processing
Received: Apr. 7, 2020
Accepted: Jun. 3, 2020
Published Online: Aug. 27, 2020
The Author Email: Ruiming Jia (jiaruiming@ncut.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20