深度學習(deep learning，DL)是指建置神經網路模型後以數據驅動的方式來擬合各種的輸入輸出的函數關係，在電腦視覺(computer vision，CV)、自然語言處理(natural language processing，NLP)、大語言模型(large language model，LLM)等領域上有許多應用。本研究是以DL建立神經網路模型搭配兩個雷射系統應用，一為超快雷射，二為適應性光學(adaptive optics system，AOS)。超快雷射是將能量集中在皮秒或飛秒等級以內激發出生物或材料的非線性光學性質，因雷射在時域上的脈衝寬度是飛秒等級而無法使用示波器等儀器量測，本研究使用空間頻域干涉術(spatial spectral interferometry，SSI)來作為量測雷射頻域相位，使用SSI拍攝干涉影像後利用Fourier filtering演算法分析出頻域相位，引入DL與AMD/Xilinx Kria SOM (system on module)中對於加速DL計算的技術來進行SSI分析時間的加速，本研究建立了F-DASI (FPGA and DNN-accelerated spatial-spectral interferometry)模型，在速度上較Fourier filtering有著13.4倍的速度提升。AOS是一種利用感知器感知目前光的變化而進行補償光學系統相位變化的一種技術，將可調變式聚焦鏡(deformable mirror，DM)擺設45度避免分光鏡(beam splitter，BS)提高雷射能量效率，再利用DL進行SWHS (Shack-Hartmann wavefront sensor)與Zernike多項式係數的分析，DL亦進行斜向入射DM的驅動器電壓向量鑑別，分別使用SHWS影像、一維聚焦點位移矩陣、三維聚焦點位移矩陣作為輸入，來訓練出可預測出正確的DM驅動器電壓向量組合的模型。

Deep learning (DL) involves building neural network models that can be trained in a data-driven manner. As a result, these models can fit various input-output functional relationships, with numerous mature applications in fields such as computer vision (CV), natural language processing (NLP), and large language models (LLM). Ultrafast lasers concentrate energy within the range of picoseconds or femtoseconds, exciting the nonlinear optical properties of biological or material substances. However, due to the laser's pulse width being at the femtosecond level, it cannot be measured with instruments like oscilloscopes. This study uses spatial spectral interferometry (SSI) to measure the laser's frequency domain phase. After capturing interference images with SSI, filtering algorithms are used to analyze the frequency domain phase. DL and AMD/Xilinx Kria SOM technology are introduced to optimize DL computations and accelerate SSI analysis time. This study established the F-DASI (FPGA and DNN-accelerated spatial-spectral interferometry) model, achieving over a 13-fold speed improvement compared to traditional filtering algorithms. Adaptive optics systems (AOS) is a technology that uses sensors to perceive changes in current light and compensate for phase variations in optical systems. This study employs a deformable mirror (DM) positioned at 45 degrees to avoid using a beam splitter (BS), thus improving laser energy efficiency. Deep learning techniques help analyze the Shack-Hartmann wavefront sensor (SHWS) and Zernike polynomial coefficients, and also determine the actuator voltage vectors for obliquely incident DM using deep learning. We trained models using SHWS images (focus matrix images), one-dimensional focus point displacement matrices, and three-dimensional focus point displacement matrices as inputs to predict the correct actuator voltage vector combinations. Finally, the accuracy of these three models was compared.

1. Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, “Gradient-based learning applied to document recognition,” Proc. IEEE 86, 2278–2324 (1998)
2. M. Göppert‐Mayer, “Elementary processes with two quantum transitions,” Ann. Phys. 18(7–8), 466–479 (2009).
3. W. Denk, J. H. Strickler, and W. W. Webb, “Two-Photon Laser Scanning Fluorescence Microscopy,” Science. 248(4951), 73–76 (1990).
4. W. R. Zipfel, R. M. Williams, and W. W. Webb,” Nonlinear magic: multiphoton microscopy in the biosciences,” Nat Biotechnol. 21(11), 1369–1377 (2003).
5. P. J T. Wang and C. Xu, “Three-photon neuronal imaging in deep mouse brain,” Optica 7(8), 947–960 (2020).
6. S.V. Plotnikov, A.M. Kenny, and S.J. Walsh, “Measurement of muscle disease by quantitative second-harmonic generation imaging,” J Biomed Opt. 13(4), 044018 (2008).
7. L. Perrin, B. Bayarmagnai, and B. Gligorijevic, “Frontiers in intravital multiphoton microscopy of cancer,” Cancer Rep. 3(1), e1192 (2020).
8. G. Bonamis, E. Audouard, C. Hönninger, J. Lopez, K. Mishchik, E. Mottay, and I. Manek-Hönninger, “Systematic study of laser ablation with GHz bursts of femtosecond pulses,” Opt. Express 28(19), 27702–27714 (2020).
9. M. Malinauskas, A. Žukauskas, S. Hasegawa, “Ultrafast laser processing of materials: from science to industry,” Light Sci Appl 5, e16133 (2016).
10. H Lin, C Wei, G Wang, H Chen, L Lin, M Ni, J Chen, S Zhuo, “Automated classification of hepatocellular carcinoma differentiation using multiphoton microscopy and deep learning,” J. Biophoton., 12(7), 2019.
11. C. Rodríguez and N. Ji, “Adaptive optical microscopy for neurobiology,” Curr. Opin. Neurobiol. 50, 83–91 (2018).
12. L. Kong and M. Cui, “In vivo neuroimaging through the highly scattering tissue via iterative multi-photon adaptive compensation technique,” Opt. Express 23(5), 6145–6150 (2015).
13. Z. Qin, S. He, C. Yang, J. S.-Y. Yung, C. Chen, C. K.-S. Leung, K. Liu, and J. Y. Qu, “Adaptive optics two-photon microscopy enables near-diffraction-limited and functional retinal imaging in vivo,” Light Sci. Appl. 9, 79 (2020).
14. M. Negro, M. Quintavalla, J. Mocci, A. G. Ciriolo, M. Devetta, R. Muradore, S. Stagira, C. Vozzi, and S. Bonora, “Fast stabilization of a high-energy ultrafast OPA with adaptive lenses,” Sci. Rep. 8, 4–9 (2018).
15. R. R. Schaller, IEEE Spectr. 34, 52 (1997).
16. K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2016), pp. 770–778.
17. J. Redmon, S. Divvala, R. Girshick, et al., “You only look once: unified, real-time object detection,” in IEEE Conference on Computer Vision and Pattern Recognition (2016).
18. H. P. Weber, “Method for Pulsewidth Measurement of Ultrashort Light Pulses Generated by Phase-Locked Lasers using Nonlinear Optics,” J. Appl. Phys. 38(5), 22312234 (1967)
19. J. A. Armstrong, “MEASUREMENT OF PICOSECOND LASER PULSE WIDTHS,” Appl. Phys. Lett. 10(1), 1618 (1967).
20. J-C. M. Diels, J. J. Fontaine, I. C. McMichael, and F. Simoni, “Control and measurement of ultrashort pulse shapes (in amplitude and phase) with femtosecond accuracy,” Appl. Opt. 24(9), 1270–1282 (1985).
21. R. Trebino and D. J. Kane, “Using phase retrieval to measure the intensity and phase of ultrashort pulses: frequency-resolved optical gating,” J. Opt. Soc. Am. A 10(5), 1101–1111 (1993).
22. R. Trebino, Frequency-Resolved Optical Gating: The Measurement of Ultrashort Laser Pulses (Kluwer Academic, 2002).
23. P. O'Shea, S. Akturk, M. Kimmel, and R. Trebino, “Practical issues in ultra-short-pulse measurements with 'GRENOUILLE',” Appl. Phys. B 79, 683-691 (2004).
24. D. Meshulach, D. Yelin, and Y. Silberberg, “Real-time spatial-spectral interference measurements of ultrashort optical pulses,” J. Opt. Soc. Am. B 14, 2095–2098 (1997).
25. P. Bowlan, P. Gabolde, A. Schreenath, K. McGresham, and R. Trebino, “Crossed-beam spectral interferometry: a simple, high-spectral-resolution method for completely characterizing complex ultrashort pulses in real time,” Opt. Express 14, 11892–11900 (2006).
26. P. Bowlan, P. Gabolde, M. A. Coughlan, R. Trebino, and R. J. Levis, “Measuring the spatiotemporal electric field of ultrashort pulses with high spatial and spectral resolution,” J. Opt. Soc. Am. B 25(6), A81–A91 (2008).
27. D. N. Fittinghoff, J. L. Bowie, J. N. Sweetser, R. T. Jennings, M. A. Krumbügel, K. W. DeLong, R. Trebino, and I. A. Walmsley, “Measurement of the intensity and phase of ultraweak, ultrashort laser pulses,” Opt. Lett. 21, 884–886 (1996).
28. M. A. Krumbügel, C. L. Ladera, K. W. DeLong, D. N. Fittinghoff, J. N. Sweetser, and R. Trebino, “Direct ultrashort-pulse intensity and phase retrieval by frequency-resolved optical gating and a computational neural network,” Opt. Lett. 21(2), 143–145 (1996).
29. R. Ziv, A. Dikopoltsev, T. Zahavy, I. Rubinstein, P. Sidorenko, O. Cohen, and M. Segev, “Deep learning reconstruction of ultrashort pulses from 2D spatial intensity patterns recorded by an all-in-line system in a single-shot,” Opt. Express 28(5), 7528–7538 (2020).
30. T. Zahavy, A. Dikopoltsev, D. Moss, G. I. Haham, O. Cohen, S. Mannor, and M. Segev, “Deep learning reconstruction of ultrashort pulses,” Optica 5(5), 666–673 (2018).
31. H. W. Babcock, “The possibility of compensating astronomical seeing,” Publ. Astron. Soc. Pac. 65, 229–236 (1953).
32. J. M. Beckers, “Adaptive optics for astronomy: principles, performance, and applications,” Annu. Rev. Astron. Astrophys. 31, 13–62 (1993).
33. P. Kern, “Adaptive optics prototype system for infrared astronomy, I: system description,” SPIE Proc. 1271, 243–251 (1990).
34. R. K. Tyson, Principles of Adaptive Optics, 2nd ed. (CRC, 1998).
35. B. C. Platt and R. V. Shack, “History and principles of Shack-Hartmann wavefront sensing,” J. Refract. Surg. 17(5), S573–S577 (2001).
36. R. Irwan and R. G. Lane, “Analysis of optimal centroid estimation applied to Shack–Hartmann sensing,” Appl. Opt. 38, 6737–6743 (1999).
37. S.-H. Baik, S.-K. Park, C.-J. Kim, and B. Cha, “A center detection algorithm for Shack–Hartmann wavefront sensor,” Opt. Las. Technol. 39(2), 262–267 (2007).
38. S.-H. Baik, S.-K. Park, C.-J. Kim, and B. Cha, “A center detection algorithm for Shack–Hartmann wavefront sensor,” Opt. Las. Technol. 39(2), 262–267 (2007).
39. W. H. Southwell, “Wave-front estimation from wave-front slope measurements,” J. Opt. Soc. Am. 70, 998–1006 (1980).
40. G.-M. Dai, “Modal wavefront reconstruction with Zernike polynomials and Karhunen-Loève functions,” J. Opt. Soc. Am. A 13, 1218–1225 (1996).
41. O. Kazasidis, S. Verpoort, and U. Wittrock, “Sensor for dynamic focus control of a deformable mirror,” Appl. Opt. 59, 5625-5630 (2020).
42. W. J. Shain, N. A. Vickers, B. B. Goldberg, T. Bifano, and J. Mertz, “Extended depth-of-field microscopy with a high-speed deformable mirror,” Opt. Lett. 42, 995–998 (2017).
43. A. Jewel, V. Akondi, and B. Vohnsen, “A direct comparison between a MEMS deformable mirror and a liquid crystal spatial light modulator in signal-based wavefront sensing,” J. Eur. Opt. Soc. 8 (2013).
44. L. Hu, S. Hu, W. Gong, and K. Si, Opt. Express 27, 33504 (2019).
45. L. Hu, S. Hu, W. Gong, and K. Si, “Deep learning assisted Shack–Hartmann wavefront sensor for direct wavefront detection,” Opt. Lett. 45(13), 3741–3744 (2020).
46. Y. L. He, Z. W. Liu, Y. Ning, J. Li, X. J. Xu, and Z. F. Jiang, Opt. Express 29, 17669 (2021)
47. S. C. Magalhães, F. V. Santos, P. Machado, A. Paulo Moreira, J. Dias, "Benchmarking edge computing devices for grape bunches and trunks detection using accelerated object detection single shot multibox deep learning models", Eng. Appl. Art. Intellig. 117, 105604 (2023).
48. THE NEURON, form: https://openbooks.lib.msu.edu/neuroscience/chapter/the-neuron/
49. Activation function table, from:https://medium.com/@shrutijadon/survey-on-activation-functions-for-deep-learning-9689331ba092
50. D. P. Kingma and J. Ba, “Adam: a method for stochastic optimization,” arXiv:1412.6980 (2014).
51. Kria K26 SOM Data Sheet (DS987), from:https://docs.amd.com/r/en-US/ds987-k26-som/Ordering-Information
52. AMD ZynqTM UltraScal+TM MPSoCs, from: https://www.amd.com/zh-tw/products/adaptive-socs-and-fpgas/soc/zynq-ultrascale-plus-mpsoc.html
53. DPUCZDX8G for Zynq UltraScale+ MPSoCs Product Guide (PG338), from:https://docs.amd.com/r/en-US/pg338-dpu?tocId=3xsG16y_QFTWvAJKHbisEw
54. Vitis AI User Guide (UG1414), from:https://docs.amd.com/r/en-US/ug1414-vitis-ai
55. P. Bowlan, P. Gabolde, A. Shreenath, K. McGresham, and R. Trebino, “Crossed-beam spectral interferometry: a simple, high-spectral-resolution method for completely characterizing complex ultrashort pulses in real time,” Opt. Express 14(24), 11892–11900 (2006).
56. Xin-Li Lee, Jui-Chi Chang, Xiang-Yu Ye, and Chia-Yuan Chang, "Field-programmable gate array and deep neural network-accelerated spatial-spectral interferometry for rapid optical dispersion analysis," Opt. Lett. 49, 1289-1292 (2024)
57. Thorlabs, introduction of Deformable Mirrors, from: https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=5056
58. Thorlabs, UV-Enhanced Aluminum Mirrors reflectance, from: https://www.thorlabs.com/newgrouppage9.cfm?objectgroup_id=12393
59. Edmund optics, specification of micro-lens array #64-482, from: https://www.edmundoptics.com.tw/p/microlens-array-10-x-10mm-500mum-pitch-05deg-divergence/19194/
60. A. Yen, “Straightforward path to Zernike polynomials”, JM3 20(02), 020501 (2021).
61. S. Ioffe and C. Szegedy, “Batch normalization: accelerating deep network training by reducing internal covariate shift,” in Proceedings of the 32nd International Conference on Machine Learning, F. Bach and D. Blei, eds., Vol. 37 of Proceedings of Machine Learning Research (PMLR, 2015), pp. 448–456.
62. Simon Tuohy and Adrian Gh. Podoleanu, "Depth-resolved wavefront aberrations using a coherence-gated Shack-Hartmann wavefront sensor," Opt. Express 18, 3458-3476 (2010)