• Chinese Optics Letters
  • Vol. 22, Issue 2, 020604 (2024)
Jiajia Zhao1, Guohui Chen1, Xuan Bi1, Wangyang Cai1..., Lei Yue1 and Ming Tang2,*|Show fewer author(s)
Author Affiliations
  • 1School of Computer and Communication Engineering, Changsha University of Science and Technology, Changsha 410114, China
  • 2Wuhan National Laboratory for Optoelectronics (WNLO) and National Engineering Laboratory for Next Generation Internet Access System, School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan 430074, China
  • show less
    DOI: 10.3788/COL202422.020604 Cite this Article Set citation alerts
    Jiajia Zhao, Guohui Chen, Xuan Bi, Wangyang Cai, Lei Yue, Ming Tang, "Fast mode decomposition for few-mode fiber based on lightweight neural network," Chin. Opt. Lett. 22, 020604 (2024) Copy Citation Text show less
    Pattern decomposition based on MobileNetV3_Light neural network.
    Fig. 1. Pattern decomposition based on MobileNetV3_Light neural network.
    Traditional convolution and depth-separable convolution.
    Fig. 2. Traditional convolution and depth-separable convolution.
    MobileNetV3_Light network structure.
    Fig. 3. MobileNetV3_Light network structure.
    MobileNetV3 block network structure diagram.
    Fig. 4. MobileNetV3 block network structure diagram.
    Test flow chart.
    Fig. 5. Test flow chart.
    Average correlations across training periods for the six model cases.
    Fig. 6. Average correlations across training periods for the six model cases.
    Simulated near-field light-field map, reconstructed near-field light-field map, residual images, and their correlation.
    Fig. 7. Simulated near-field light-field map, reconstructed near-field light-field map, residual images, and their correlation.
    Relation between the mode number and correlation.
    Fig. 8. Relation between the mode number and correlation.
    Simulated and reconstructed images under the influence of different intensities of noise and their correlation.
    Fig. 9. Simulated and reconstructed images under the influence of different intensities of noise and their correlation.
    InputOperatorExp size#outSENLs
    2242 × 3Conv2d8HS2
    1122 × 8Bneck, 3 × 31616RE2
    562 × 16Bneck, 3 × 37224RE2
    282 × 24Bneck, 3 × 38824RE1
    282 × 24Bneck, 5 × 59640HS2
    142 × 40Bneck, 5 × 524040HS1
    142 × 40Bneck, 5 × 512048HS1
    142 × 48Bneck, 5 × 514448HS1
    142 × 112Bneck, 5 × 528896HS2
    72 × 96Bneck, 5 × 557696HS1
    72 × 96Conv2d, 1 × 1576HS1
    72 × 576AvgPool, 7 × 71
    12 × 576Conv2d, 1 × 1, NBN1024HS1
    12 × 1024Conv2d, 1 × 1, NBNK1
    Table 1. Detailed Parameter Settings of MobileNetV3_Light Network Model Structure
     Δp1¯Δp2¯Δp3¯Δp4¯Δp5¯Δp6¯
    Average weights error0.47%0.48%0.42%0.48%0.53%0.55%
    Table 2. Average Error of the Six Model Weights
     Δθ1¯Δθ2¯Δθ3¯Δθ4¯Δθ5¯
    Average weights error0.47%0.48%0.42%0.48%0.53%
    Table 3. Average Error of the Relative Phase of the Six Modes
     T1T2T3T4
    Predicting model weight and phase267.5 min2.41 s3.86 s36.24 s
    Table 4. Time Spent in Different Phases of Testing
     MobileNetV3_LightMoblieNetV2XceptionResnet50VGG-16
    Parameters2.5 × 1063.4 × 10622.85 × 10625.56 × 106138.36 × 106
    Mode size6.5 MB14.2 MB88 MB98 MB528 MB
    Table 5. Parameter Size of Different Neural Network Models
    Jiajia Zhao, Guohui Chen, Xuan Bi, Wangyang Cai, Lei Yue, Ming Tang, "Fast mode decomposition for few-mode fiber based on lightweight neural network," Chin. Opt. Lett. 22, 020604 (2024)
    Download Citation