Lensless complex amplitude demodulation based on deep learning in holographic data storage

Hao, Jianying; Lin, Xiao; Lin, Yongkun; Chen, Mingyong; Chen, Ruixian; Situ, Guohai; Horimai, Hideyoshi; Tan, Xiaodi

doi:10.29026/oea.2023.220157

Cited by 29 publications

(10 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The main preparation methods of PQ/PMMA photopolymer materials include solvent common-method and thermopolymerization method, and this paper adopts the method of thermopolymerization [11]. In the HDS system, since the phase information cannot be directly detected by the detector, the phase reconstruction method is very important [10]. This paper uses a phase retrieval method based on deep learning for holographic data storage to explore the influence of different numbers of experimental images on the accuracy of phase reconstruction.…”

Section: Methodsmentioning

confidence: 99%

Phase retrieval based on deep learning in PQ/PMMA material

Zheng,

Hao

et al. 2023

ODS 2023: Industrial Optical Devices and Systems

Self Cite

View full text Add to dashboard Cite

Holographic data storage (HDS), a three-dimensional volume storage technology, is becoming a strong candidate for huge data storage due to its advantages of higher storage density, faster data transfer rate, and longer life. In this paper, we use phenanthraquinone doped polymethyl methacrylate (PQ/PMMA) photopolymer to record the phase data page, and use deep learning to recover the phase. The experimental setup used 532 nm laser wavelength. The phase data page uploaded on the SLM are made up of 4-gray level random phase coded (π/6,2π/3, π,3π/2). Every pixel pitch of SLM is 20 μm. Every phase data is described by some oversampling display such as 8*8 pixels on the SLM. And, the CMOS captured phase data pages using PQ/PMMA photopolymers. We input 4439 randomly generated phase data pages into the experimental system. 4439 phase data pages after rotating 90°, rotating 180°, rotating 270°, upside down, flipped left to right, expanded to 6 times, a total of 26634 phase data pages. Among them, 90 % is used as the training set to optimize the network, 10 % is used as the testing set to verify the generalization ability of the trained neural network. We trained the U-net 50 epoch, and when we got the predicted phase data. And the bit error rate (BER) of the reconstructed image using PQ/PMMA photopolymers were measured, and we found that, different numbers experimental images has different BER. So, deep learning can effectively reduce BER of phase data page by using PQ/PMMA photopolymers.

show abstract

Section: Methodsmentioning

confidence: 99%

Phase retrieval based on deep learning in PQ/PMMA material

Zheng,

Hao

et al. 2023

ODS 2023: Industrial Optical Devices and Systems

Self Cite

View full text Add to dashboard Cite

show abstract

“…These approaches usually depend on multi-DOFs optimization, including the physical dimensions and orientation angles of the individual meta-atom, or even the relative position between neighboring elements in the array. Although the emergence of deep learning techniques [ 34 , 35 , 36 ] demonstrates great potential in various fields, including metasurface, even complex-amplitude modulation and demodulation, the multi-DOFs optimization procedure requires a lot of computing resources and is time-consuming for large-scale meta-device applications. On the other hand, chiral structures are exploited to achieve spin-decoupled full-range phase modulation by combing the AA phase with the PB phase [ 28 , 37 ].…”

Section: Concept and Meta-atom Designmentioning

confidence: 99%

Full Complex-Amplitude Modulation of Surface Waves Based on Spin-Decoupled Metasurface

Quan

Xie

et al. 2023

Micromachines

View full text Add to dashboard Cite

This work proposes a method for surface wave (SW) coupling along with flexible complex amplitude modulation of its wavefront. The linearly polarized incident plane wave is coupled into the surface mode with complex wavefront by exploiting the spin-decouple nature of a reflective chiral meta-atom. As verification, two kinds of metasurface couplers are designed. The first kind contains two examples for SW airy beam generation with and without deflection under linearly polarized illumination, respectively. The second kind is a bi-functional device capable of SW focusing under left-handed circularly polarized illumination, and propagating wave deflection under right-handed circularly polarized illumination, respectively, to verify the fundamental spin-decoupled character. Simulated and experimental results are in good agreement. We believe that this method provides a flexible approach for complex SW applications in integrated optics, optical sensing, and other related fields.

show abstract

“…But most of the time, generating features and feature selection are equally experience-consuming and very unfriendly to those who lack domain knowledge. With the continuous improvement in computational power and the enhancement of material databases [10], deep learning methods based on artificial neural networks (ANN) have gained increasing popularity among researchers [11][12][13]. ANN can bypass the complex feature engineering process.…”

Section: Introductionmentioning

confidence: 99%

Predicting the properties of perovskite materials by improved compositionally restricted attention-based networks and explainable machine learning

Hui,

Wang,

Wang

et al. 2024

J. Phys. D: Appl. Phys.

View full text Add to dashboard Cite

Understanding the unique properties of perovskite materials is crucial in advancing solar energy technologies. Factors like heat of formation and bandgap significantly influence the light absorption capability and stability of perovskite solar cells. However, it is time-consuming and labor-intensive to obtain the properties of perovskites using traditional experimental or high-throughput computational methods. As a prospective method, machine learning can find regularities in the given training data and give accurate prediction results. In this article, we use deep learning models based on attention mechanisms and elemental features to predict the heat of formation and bandgap of perovskite materials. Random Forest and Gradient Boosted Regression Tree models have also been used for interpretable predictions of properties. The compositionally restricted attention-based network was improved by introducing a densely connected network and optimizing the network structure to increase data processing capabilities. The experiment results show that the mean absolute errors of the heat of formation and bandgap on the test sets are decreased by 5.77% and 3.37% respectively. The optimized model also shows better performance when used for classification tasks. In addition, we use the Gradient Boosting Regression Tree model and the Shapley Additive Explanations tool to conduct an interpretable analysis, explaining the impact of different features on the predictions of the properties.

show abstract

Lensless complex amplitude demodulation based on deep learning in holographic data storage

Cited by 29 publications

References 32 publications

Phase retrieval based on deep learning in PQ/PMMA material

Phase retrieval based on deep learning in PQ/PMMA material

Full Complex-Amplitude Modulation of Surface Waves Based on Spin-Decoupled Metasurface

Predicting the properties of perovskite materials by improved compositionally restricted attention-based networks and explainable machine learning

Contact Info

Product

Resources

About