First steps into coherent object classification using convolutional deep diffractive neural networks
-
Christian Eder
Mr. Christian Eder is a doctorate student at Aalen University in Germany. He works as a scientific staff member at the Center for Optical Technologies in the field of micro- and nano-structuring. Christian Eder achieved his Bachelor’s degree (B. Eng.) in 2018 at Aalen University for the investigation of manufacturing organic light-emitting diodes using inkjet printing. In 2020 he achieved his Master’s degree (M. Sc.) for research in the field of numerical diffraction simulation, also in at Center for Optical Technologies in Aalen. His current research is in the field of photonic neural computing with diffractive optical networks.
und
Andreas Heinrich
Dr. Andreas Heinrich is a full professor for optical technologies at Aalen University, Germany and head of the Center for Optical Technologies. Since 2013 he and his group focus not only on the development of new concepts for complex optical components, but also on the development of new additive manufacturing methods especially for the manufacturing of optical components. Dr. Andreas Heinrich got his Diploma in Physics at the Technical University of Munich, Germany in 1999 and in 2001 his PhD in the field of solid-state physics from the Technical University of Munich, as well. From 2001 until 2006 he did his habilitation at the University of Augsburg, Germany and at the Florida State University, USA. From 2007–2013 he headed a research group in the field of applied optics at Carl Zeiss, Germany.
Abstract
As artificial intelligence and deep learning becomes more important, new approaches for photonic neural computing arise. We investigate the concept of deep diffractive neural networks. First proposed in 2018, deep diffractive neural network operate passively, using coherent images and diffractive optics to do image-to-image regression and object classification. In this article we shortly review current approaches, give a brief introduction into the mathematical description of such diffractive networks using the Angular Spectrum method and show the first results of our own developments of convolutional diffractive networks with an experimental accuracy of approximately 84 %. The objective of this article is to give an introduction into the field of optical computing with neural networks using diffraction and free-space propagation of light.
Zusammenfassung
Mit steigendem Interesse an künstlicher Intelligenz und Deep Learning, werden auch neue Ansätze im Bereich der optischen neuronalen Netzwerke untersucht. In diesem Artikel wollen wir einen Einblick in die Welt der diffraktiven neuronalen Netzwerke werfen. Der 2018 von Lin et al. vorgestellte Ansatz eines diffraktiven neuronalen Netzwerks arbeitet passiv und unter der Verwendung von kohärenten beleuchteten Bildern und der Beugung von Licht um Bild-zu-Bild-Regression oder Objektklassifizierung optisch zu realisieren. Wir beschreiben kurz den bisherigen Ansatz, geben einen Einblick in die mathematische Beschreibung dieser Netzwerke, mittels der Angular-Spectrum-Methode, und stellen die ersten theoretischen Ergebnisse unseres Ansatzes eines diffraktiven Falungsnetzwerkes vor. Unser diffraktives Faltungsnetzwerk erreicht eine Genauigkeit von ca. 84 % beim Erkennen handgeschriebener Zahlen. Mit diesem Artikel wollen wir eine Einführung in das Thema der diffraktiven neuralen Netzwerke geben.
About the authors
Mr. Christian Eder is a doctorate student at Aalen University in Germany. He works as a scientific staff member at the Center for Optical Technologies in the field of micro- and nano-structuring. Christian Eder achieved his Bachelor’s degree (B. Eng.) in 2018 at Aalen University for the investigation of manufacturing organic light-emitting diodes using inkjet printing. In 2020 he achieved his Master’s degree (M. Sc.) for research in the field of numerical diffraction simulation, also in at Center for Optical Technologies in Aalen. His current research is in the field of photonic neural computing with diffractive optical networks.
Dr. Andreas Heinrich is a full professor for optical technologies at Aalen University, Germany and head of the Center for Optical Technologies. Since 2013 he and his group focus not only on the development of new concepts for complex optical components, but also on the development of new additive manufacturing methods especially for the manufacturing of optical components. Dr. Andreas Heinrich got his Diploma in Physics at the Technical University of Munich, Germany in 1999 and in 2001 his PhD in the field of solid-state physics from the Technical University of Munich, as well. From 2001 until 2006 he did his habilitation at the University of Augsburg, Germany and at the Florida State University, USA. From 2007–2013 he headed a research group in the field of applied optics at Carl Zeiss, Germany.
References
1. LeCun, Y., Haffner, P., Bottou, L., Bengio, Y.: Object Recognition with Gradient-Based Learning. In: Shape, Contour and Grouping in Computer Vision, pp. 319–345. Springer, Berlin, Heidelberg (1999).10.1007/3-540-46805-6_19Suche in Google Scholar
2. Fuyong, X., Yuanpu, X., Hai, S., Fujun, L., Lin, Y.: Deep Learning in Microscopy Image Analysis: A Survey. IEEE transactions on neural networks and learning systems (2018). https://doi.org/10.1109/TNNLS.2017.2766168.10.1109/TNNLS.2017.2766168Suche in Google Scholar PubMed
3. Litjens, G., Kooi, T., Bejnordi, B.E., Setio, A.A.A., Ciompi, F., Ghafoorian, M., van der Laak, J.A.W.M., van Ginneken, B., Sánchez, C.I.: A survey on deep learning in medical image analysis. Medical image analysis (2017). https://doi.org/10.1016/j.media.2017.07.005.10.1016/j.media.2017.07.005Suche in Google Scholar PubMed
4. Protopapadakis, E., Voulodimos, A., Doulamis, A., Doulamis, N., Stathaki, T.: Automatic crack detection for tunnel inspection using deep learning and heuristic image post-processing. Appl Intell (2019). https://doi.org/10.1007/s10489-018-01396-y.10.1007/s10489-018-01396-ySuche in Google Scholar
5. Tian, C., Fei, L., Zheng, W., Xu, Y., Zuo, W., Lin, C.-W.: Deep learning on image denoising: An overview. Neural networks: the official journal of the International Neural Network Society (2020). https://doi.org/10.1016/j.neunet.2020.07.025.10.1016/j.neunet.2020.07.025Suche in Google Scholar PubMed
6. Tompson, J., Goroshin, R., Jain, A., LeCun, Y., Bregler, C.: Efficient object localization using Convolutional Networks. In: 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE (2015). https://doi.org/10.1109/cvpr.2015.7298664.10.1109/CVPR.2015.7298664Suche in Google Scholar
7. Minaee, S., Boykov, Y.Y., Porikli, F., Plaza, A.J., Kehtarnavaz, N., Terzopoulos, D.: Image Segmentation Using Deep Learning: A Survey. IEEE Trans. Pattern Anal. Mach. Intell. (2021). https://doi.org/10.1109/tpami.2021.3059968.10.1109/TPAMI.2021.3059968Suche in Google Scholar PubMed
8. LeCun, Y., Bengio, Y., Hinton, G.: Deep learning. Nature (2015). https://doi.org/10.1038/nature14539.10.1038/nature14539Suche in Google Scholar PubMed
9. Burgess, J.: RTX on—The NVIDIA Turing GPU. IEEE Micro (2020). https://doi.org/10.1109/mm.2020.2971677.10.1109/HOTCHIPS.2019.8875651Suche in Google Scholar
10. Tait, A.N., Nahmias, M.A., Shastri, B.J., Prucnal, P.R.: Chapter 8 Photonic Neuromorphic Signal Processing and Computing. In: (2015).Suche in Google Scholar
11. Esser, S.K., Merolla, P.A., Arthur, J.V., Cassidy, A.S., Appuswamy, R., Andreopoulos, A., Berg, D.J., McKinstry, J.L., Melano, T., Barch, D.R., Di Nolfo, C., Datta, P., Amir, A., Taba, B., Flickner, M.D., Modha, D.S.: Convolutional networks for fast, energy-efficient neuromorphic computing. PNAS (2016). https://doi.org/10.1073/pnas.1604850113.10.1073/pnas.1604850113Suche in Google Scholar PubMed PubMed Central
12. Feldmann, J., Youngblood, N., Karpov, M., Gehring, H., Li, X., Stappers, M., Le Gallo, M., Fu, X., Lukashchuk, A., Raja, A.S., Liu, J., Wright, C.D., Sebastian, A., Kippenberg, T.J., Pernice, W.H.P., Bhaskaran, H.: Parallel convolutional processing using an integrated photonic tensor core. Nature (2021). https://doi.org/10.1038/s41586-020-03070-1.10.1038/s41586-020-03070-1Suche in Google Scholar
13. Lin, X., Rivenson, Y., Yardimci, N.T., Veli, M., Luo, Y., Jarrahi, M., Ozcan, A.: All-optical machine learning using diffractive deep neural networks. Science (New York, N.Y.) (2018). https://doi.org/10.1126/science.aat8084.10.1126/science.aat8084Suche in Google Scholar
14. de Marinis, L., Cococcioni, M., Castoldi, P., Andriolli, N.: Photonic Neural Networks: A Survey. IEEE Access (2019). https://doi.org/10.1109/ACCESS.2019.2957245.10.1109/ACCESS.2019.2957245Suche in Google Scholar
15. Aggarwal, C.C.: Neural networks and deep learning. A textbook / Charu C. Aggarwal. Springer, Cham, Switzerland (2018).10.1007/978-3-319-94463-0Suche in Google Scholar
16. Janocha, K., Czarnecki, W.M.: On Loss Functions for Deep Neural Networks in Classification. http://arxiv.org/pdf/1702.05659v1 (2017).10.4467/20838476SI.16.004.6185Suche in Google Scholar
17. Yan, T., Wu, J., Zhou, T., Xie, H., Xu, F., Fan, J., Fang, L., Lin, X., Dai, Q.: Fourier-space Diffractive Deep Neural Network. Physical review letters (2019). https://doi.org/10.1103/PhysRevLett.123.023901.10.1103/PhysRevLett.123.023901Suche in Google Scholar
18. Zhou, T., Fang, L., Yan, T., Wu, J., Li, Y., Fan, J., Wu, H., Lin, X., Dai, Q.: In situ optical backpropagation training of diffractive optical neural networks. Photon. Res. (2020). https://doi.org/10.1364/PRJ.389553.10.1364/PRJ.389553Suche in Google Scholar
19. Matsushima, K., Shimobaba, T.: Band-limited angular spectrum method for numerical simulation of free-space propagation in far and near fields. Optics express (2009). https://doi.org/10.1364/OE.17.019662.10.1364/OE.17.019662Suche in Google Scholar
20. He, G.: Optical phase conjugation: principles, techniques, and applications. Progress in Quantum Electronics (2002). https://doi.org/10.1016/S0079-6727(02)00004-6.10.1016/S0079-6727(02)00004-6Suche in Google Scholar
21. Zuo, Y., Li, B., Zhao, Y., Jiang, Y., Chen, Y.-C., Chen, P., Jo, G.-B., Liu, J., Du, S.: All-optical neural network with nonlinear activation functions. Optica (2019). https://doi.org/10.1364/OPTICA.6.001132.10.14711/thesis-991012994404403412Suche in Google Scholar
22. Chang, J., Sitzmann, V., Dun, X., Heidrich, W., Wetzstein, G.: Hybrid optical-electronic convolutional neural networks with optimized diffractive optics for image classification. Sci Rep (2018). https://doi.org/10.1038/s41598-018-30619-y.10.1038/s41598-018-30619-ySuche in Google Scholar PubMed PubMed Central
© 2022 Walter de Gruyter GmbH, Berlin/Boston
Artikel in diesem Heft
- Frontmatter
- Editorial
- Multidimensionale optische Messtechnik
- Beiträge
- The coherence function and its information content for optical metrology
- Ereignisbasierte Weißlichtinterferometrie (eCSI)
- First steps into coherent object classification using convolutional deep diffractive neural networks
- Statistical analysis of phase values for the determination of step heights in multi-wavelength interferometry
- Abbildende Müller-Matrix-Ellipsometrie für die Charakterisierung vereinzelter Nanostrukturen
- Light-sheet fluorescence microscopic probing of silicate materials
- Calibration of focus-adjustable cameras for triangulation sensors by model-based camera parameter estimation
Artikel in diesem Heft
- Frontmatter
- Editorial
- Multidimensionale optische Messtechnik
- Beiträge
- The coherence function and its information content for optical metrology
- Ereignisbasierte Weißlichtinterferometrie (eCSI)
- First steps into coherent object classification using convolutional deep diffractive neural networks
- Statistical analysis of phase values for the determination of step heights in multi-wavelength interferometry
- Abbildende Müller-Matrix-Ellipsometrie für die Charakterisierung vereinzelter Nanostrukturen
- Light-sheet fluorescence microscopic probing of silicate materials
- Calibration of focus-adjustable cameras for triangulation sensors by model-based camera parameter estimation
