UTIRnet

所属分类:数值算法/人工智能
开发工具:matlab
文件大小:0KB
下载次数:2
上传日期:2024-04-08 07:28:49
上 传 者sh-1993
说明:  数字同轴全息显微术中孪晶图像去除的神经网络方法
(Neural network for twin-image removal in digital in-line holographic microscopy)

文件列表:
Holograms/
Networks/
AS_propagate_p.m
GenerateDataset.m
GenerateHologram.m
LICENSE
NetworkArchitecture.mat
UTIRnetReconstruction.m
main_example.m

# UTIRnet UTIRnet is a supervised universal convolutional neural network for twin-image effect removal in digital in-line holographic microscopy (DIHM) # Contents and codes - main_example.m - main code with examples how to: (1) generate network training data, (2) train UTIRnet and (3) reconstruct holograms with UTIRnet
- AS_propagate_p.m - function for optical field propagation with angular spectrum (AS) method
- GenerateDataset.m - function for generating a whole dataset for UTIRnet training
- GenerateHologram.m - function for generating a single hologram and a pair of input-target images for network training
- NetworkArchitecture.m - function for creating CNN architecture
- UTIRnetReconstruction.m - function for reconstructing holograms with the UTIRnet
- ./Holograms - directory where experimental holograms may be stored (see Experimental data section)
- ./Networks - directory with two trained UTIRnets that we employed in our article (see Cite as section)
# How does it work Follow the steps in main_example.m code to generate synthetic training data and then to train UTIRnet network for a specified system parameters (wavelength, pixel size, magnification, sample-focus plane distance (or sample-camera distance in lensless DIHM system)). Then, generated network (composed from CNN_A and CNN_P networks) along with AS propagation may be used to reconstruct holograms without twin-image effect. # Experimental data Our experimental data (holograms and reference reconstructions) may be found at:
M. Rogalski, P. Arcab, L. Stanaszek, V. Micó, C. Zuo, and M. Trusiak, “Physics-driven universal twin-image removal network for digital in-line holographic microscopy - dataset,” Jun. 2023, doi: 10.5281/ZENODO.8059636.
https://zenodo.org/record/8059636 # Cite as M. Rogalski, P. Arcab, L. Stanaszek, V. Micó, C. Zuo, and M. Trusiak, “Physics-driven universal twin-image removal network for digital in-line holographic microscopy,” Opt. Express, vol. 32, no. 1, p. 742, Jan. 2024, doi: 10.1364/OE.505440. # Created by Mikoaj Rogalski,
mikolaj.rogalski.dokt@pw.edu.pl
Institute of Micromechanics and Photonics,
Warsaw University of Technology, 02-525 Warsaw, Poland

近期下载者

相关文件


收藏者