Modal Decomposition of Hermite-Gaussian Beams Through Deep Learning

shu Peng; Xulian Guo; Tianbao Ma; Kui Liu; Jiangrui Gao

doi:10.3788/LOP240667

Laser & Optoelectronics Progress, Volume. 61, Issue 23, 2306003(2024)

Modal Decomposition of Hermite-Gaussian Beams Through Deep Learning

shu Peng¹, Xulian Guo¹, Tianbao Ma¹, Kui Liu^1,2、*, and Jiangrui Gao^1,2

¹State Key Laboratory of Quantum Optics and Quantum Optics Devices, Institute of Opto-Electronics, Shanxi University, Taiyuan 030006, Shanxi , China

²Collaborative Innovation Center of Extreme Optics, Shanxi University, Taiyuan 030006, Shanxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(8)

Fig. 1. 6 modes of light intensity. (a) HG₀₀ mode; (b) HG₁₀ mode; (c) HG₀₁ mode; (d) HG₂₀ mode; (e) HG₁₁ mode; (f) HG₀₂ mode

Download full size

Fig. 2. Experimental device diagram of generating 6 modes of light intensity diagram

Download full size

Fig. 3. Mode decomposition flowchart of network model. (a) An overall flowchart of light intensity diagram pattern decomposition; (b) diagram of overall structure of Swin Transformer network

Download full size

Fig. 4. Theoretical light intensity graph training result graph. (a) Trend of weight MSE; (b) trend of phase MSE

Download full size

Fig. 5. Experimental light intensity map training results. (a) Trend of weighted MSE; (b) trend of phase MSE

Download full size

Fig. 6. Light intensity diagram, phase diagram, and weight phase comparison diagrams of test sample. (a) Experimental light intensity diagram; (b) theoretical light intensity diagram; (c) theoretical phase diagram; (d) reconstructed theoretical light intensity diagram; (e) reconstructed theoretical phase diagram; (f) weight comparison diagram of true and predicted values; (g) phase comparison diagram of true and predicted values

Download full size

Table 1. Comparison of prediction error between Swin Transformer and VGG16

View table

Table 1. Comparison of prediction error between Swin Transformer and VGG16

Error item	$Δ ρ_{00}^{2}$	$Δ ρ_{10}^{2}$	$Δ ρ_{01}^{2}$	$Δ ρ_{20}^{2}$	$Δ ρ_{11}^{2}$	$Δ ρ_{02}^{2}$
Swin Transformer	0.01880	0.01643	0.01682	0.00356	0.00322	0.00320
VGG16	0.03550	0.03550	0.03600	0.01850	0.03240	0.01640
Error item	$Δ θ_{10}$	$Δ θ_{01}$	$Δ θ_{20}$	$Δ θ_{11}$	$Δ θ_{02}$	E_PE
Swin Transformer	0.01400	0.01414	0.01612	0.01703	0.01750	0.01670
VGG16	0.03920	0.03770	0.04390	0.04410	0.04420	0.01330

Table 2. Error in weight and phase of experimental data on the validation set
View table
Table 2. Error in weight and phase of experimental data on the validation set
Error item $Δ ρ_{00}^{2}$ $Δ ρ_{10}^{2}$ $Δ ρ_{01}^{2}$ $Δ ρ_{20}^{2}$ $Δ ρ_{11}^{2}$ $Δ ρ_{02}^{2}$
Averaged 0.0520 0.0562 0.0560 0.0447 0.0483 0.041
Error item $Δ θ_{10}$ $Δ θ_{01}$ $Δ θ_{20}$ $Δ θ_{11}$ $Δ θ_{02}$ $E_{P E}$
Averaged 0.1597 0.1701 0.1622 0.1608 0.1577 0.1317

Tools

Get Citation

Copy Citation Text

shu Peng, Xulian Guo, Tianbao Ma, Kui Liu, Jiangrui Gao. Modal Decomposition of Hermite-Gaussian Beams Through Deep Learning[J]. Laser & Optoelectronics Progress, 2024, 61(23): 2306003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites