Degenerated mode decomposition with convolutional neural network for few-mode fibers

Yan, Baorui; Zhang, Jianyong; Wang, Muguang; Jiang, Youchao; Mi, Shuchao

doi:10.1016/j.optlastec.2022.108287

Cited by 14 publications

(9 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It can be seen from Table 5 that the number of parameters of our proposed MobileNetV3_Light network model is only 2.5 million and the size of the network model is 6.5 MB. Compared to the neural network model size of some of the proposed pattern decompositions [36,37] , the decomposition scheme proposed in this paper has obvious advantages. Currently proposed neural network methods for MD generally have the problem of large model size, which we avoid better by designing a lightweight neural network.…”

Section: Resultsmentioning

confidence: 99%

Fast mode decomposition for few-mode fiber based on lightweight neural network

Zhao,

Chen,

et al. 2024

中国光学快报

View full text Add to dashboard Cite

In this paper, we present a fast mode decomposition method for few-mode fibers, utilizing a lightweight neural network called MobileNetV3-Light. This method can quickly and accurately predict the amplitude and phase information of different modes, enabling us to fully characterize the optical field without the need for expensive experimental equipment. We train the MobileNetV3-Light using simulated near-field optical field maps, and evaluate its performance using both simulated and reconstructed near-field optical field maps. To validate the effectiveness of this method, we conduct mode decomposition experiments on a few-mode fiber supporting six linear polarization (LP) modes (LP01, LP11e, LP11o, LP21e, LP21o, LP02). The results demonstrate a remarkable average correlation of 0.9995 between our simulated and reconstructed near-field lightfield maps. And the mode decomposition speed is about 6 ms per frame, indicating its powerful real-time processing capability. In addition, the proposed network model is compact, with a size of only 6.5 MB, making it well suited for deployment on portable mobile devices.

show abstract

Section: Resultsmentioning

confidence: 99%

Fast mode decomposition for few-mode fiber based on lightweight neural network

Zhao,

Chen,

et al. 2024

中国光学快报

View full text Add to dashboard Cite

show abstract

“…The model consists of convolutional, pooling and fully connected layers, as well as an output layer (softmax activation layer for 5 categories). The model was trained for images of 150x150 pixel size for a total of 10 epochs, using the adam optimizer and categorical cross-entropy as loss functions [9,10].…”

Section: Modelingmentioning

confidence: 99%

Research on apple detection and maturity assessment based on computer vision technology

Zheng,

Zhang,

Wang

2024

HSET

View full text Add to dashboard Cite

The study is based on computer vision technology for apple detection and maturity assessment. Firstly, the YOLOv7 algorithm is used for target detection of apples, and the number of apples in the image is counted to generate a distribution histogram of the number of apples. Then, the position of each apple is detected by the YOLOv7 algorithm and a 2D scatter plot of the geometric coordinates of the apples is drawn. Next, apple foreground was extracted by interactive ROI tagging and GrabCut algorithms, and a mathematical model was developed to assess the maturity of apples based on the histogram analysis of apple colours in HSV colour space model. In addition, the Faster R-CNN model is used to detect apples and estimate the quality of apples based on the 2D area of apples to generate a distribution histogram of apple quality. Finally, a convolutional neural network is used to build a fruit recognition model and draw a distribution histogram of apple image ID numbers. This study provides computer vision technical support for apple picking robots to improve apple production efficiency.

show abstract

“…The family of non-iterative approaches includes a variety of machine-learning-based algorithms, primarily based on deep convolutional neural networks (CNNs) [18,[27][28][29]. Deep unsupervised learning was also applied [30], which demonstrated MD in 10-mode fibers with low accuracy.…”

Section: Introductionmentioning

confidence: 99%

Sparse intensity sampling for ultrafast full-field reconstruction in low-dimensional photonic systems

Manuylovich

2024

Preprint

View full text Add to dashboard Cite

Exploiting light’s phase dimension opened multiple avenues in photonics, from phase-contrast microscopy to coherent optical communications. Phase-sensitive measurements usually utilize interferometric techniques to extract the optical phase. However, in some cases, when the feature space of an electromagnetic field is highly degenerate or inherently low dimensional, the vast majority of field distribution parameters, including wavefront, can be extracted from intensity measurements only. But even the fastest of previously published intensity-only methods have too high a computational complexity to be applicable at the speeds used in optical communications and, most importantly, require data from CCD cameras, which are generally very slow. This paper shows how a few intensity measurements, taken from properly placed photodetectors, can be used to reconstruct the complex-valued field fully in systems with low-dimensional feature space. The presented method allows full-field characterization in few-mode fibers and does not employ a reference beam. It is 3 orders of magnitude faster than the fastest previously published result and uses 3 orders of magnitude fewer photodetectors, allowing retrieval of mode amplitudes and phases relative to the fundamental mode using only several telecom-grade photodetectors. This opens the avenue for ultra-fast applications of intensity-only mode decomposition methods, including the pulse-to-pulse beam complete characterization of lasers with spatio-temporal modelocking, providing an essential tool for experimental exploration of the modal dynamics in such systems. It can also be applied to ultra-fast sensing in few-mode fibers and for coherent mode division-multiplexed receivers using quadratic detectors only.

show abstract

Degenerated mode decomposition with convolutional neural network for few-mode fibers

Cited by 14 publications

References 36 publications

Fast mode decomposition for few-mode fiber based on lightweight neural network

Fast mode decomposition for few-mode fiber based on lightweight neural network

Research on apple detection and maturity assessment based on computer vision technology

Sparse intensity sampling for ultrafast full-field reconstruction in low-dimensional photonic systems

Contact Info

Product

Resources

About