“Deep learning on galaxy morphology profile”的版本间差异

来自cluster
跳到导航 跳到搜索
无编辑摘要
无编辑摘要
 
（未显示同一用户的30个中间版本）
第1行： 第1行：

本项目主要根据星系形态将星系图片拟合至其形态参数。
==Introduction==
第一项工作旨在复现文章《Deep learning for galaxy surface brightness profile fitting》中用星系信息对星系的参数进行拟合任务，星系形态参数包含星系的星等、星系的轮廓Sersic指数、星系的半光半径和星系的轴比。
*This project focuses on fitting galaxy pictures to their morphological parameters based on galaxy morphology.
引文提出了一种基于二维卷积神经网络的二维光度星系轮廓建模新方法：DeepLeGATo，以实现对星系形态的自动化和快速分析。

首先用GalSim，根据随机参数生成相关星系的信息文件。因为深度学习需要大量的图像和参数样本，而图像手工标注的成本太高；且对进行注释人员专业知识要求较高。用参数生成数据集为较方便的实验方法。
==Relate Work==
用GalSim软件生成星系数据时，因为真实星系数据维度大小不一，即将生成的数据的维度大小也设定为随机。用该python库进行生成50000个星系信息，将其以8：1：1的比例分割为训练集、测试集和验证集。为处理不同大小的星系图像，本文采用标准大小为128*128大小的二维数据，将形状不同于128*128大小的数据，剪切大于128*128像素的星系图像的中心，填充小于128*128像素的图像的边缘。
*The first work aims to reproduce the task of fitting the parameters of galaxies with galaxy information in the article "Deep learning for galaxy surface brightness profile fitting". The galaxy morphology parameters contain the magnitude of galaxies, the profile Sersic index of galaxies, the half-light radius of galaxies, and the axis ratio of galaxies.
*The cited paper presents a new method for modeling 2D photometric galaxy profiles based on 2D convolutional neural networks: DeepLeGATo for automated and fast analysis of galaxy morphology.
实验中构建与原文结构相似的神经网络结构，所用的神经网络结构如下图所示：

⚫
[[File:111.png|500px|center]]
==Generate Method==
以上网络用二维卷积神经网络将星系数据对一个参数进行拟合。以下是训练单参数——星等的训练结果：
===First Method===
[[File:train_loss.png|500px|center]] [[File:trian_r2.png|500px|center]]
[[File:test_loss.png|500px|center]] [[File:test_r2.png|500px|center]]
[[File:111.png|500px|right|jumengting]]
*The neural network structure used in the experiments to construct a neural network structure similar to the original text is shown in the following figure.
第二项工作将数据拟合至多个参数。
*Firstly, GalSim is used to generate information files of relevant galaxies based on random parameters. Because deep learning requires a large number of images and parameter samples, the cost of manual annotation of images is too high; and the expertise of the person conducting the annotation is required. Generating datasets with parameters is a more convenient experimental method.
*When generating galaxy data with GalSim software, the dimension size of the generated data is also set to random because the dimension size of the real galaxy data varies. The python library is used to generate 50,000 galaxy information, which is partitioned into a training set, a test set, and a validation set in the ratio of 8:1:1. In order to handle galaxy images of different sizes, this paper uses two-dimensional data with a standard size of 128*128 size, which will be different in shape from 128*128 size, cut the center of galaxy images larger than 128*128 pixels, and fill the edges of images smaller than 128*128 pixels.
===Second NN===
*The second task fits the data to multiple parameters.
⚫
[[File:Nn_111.png|500px|right|jumengting]]

==Result==
===Generate data First Method===
*The above network fits the galaxy data to one parameter using a 2D convolutional neural network. The following are the training results for training a single parameter - star magnitude.
[[File:result_111.png|500px|right|jumengting]]
===Generate data Second Method===
*Galsim result.
===Ground-Truth data Second Method===
*CANDELS

==Else==
Waiting...

2022年10月1日 (六) 08:15的最新版本

Introduction

  • This project focuses on fitting galaxy pictures to their morphological parameters based on galaxy morphology.

Relate Work

  • The first work aims to reproduce the task of fitting the parameters of galaxies with galaxy information in the article "Deep learning for galaxy surface brightness profile fitting". The galaxy morphology parameters contain the magnitude of galaxies, the profile Sersic index of galaxies, the half-light radius of galaxies, and the axis ratio of galaxies.
  • The cited paper presents a new method for modeling 2D photometric galaxy profiles based on 2D convolutional neural networks: DeepLeGATo for automated and fast analysis of galaxy morphology.

Generate Method

First Method

jumengting
  • The neural network structure used in the experiments to construct a neural network structure similar to the original text is shown in the following figure.
  • Firstly, GalSim is used to generate information files of relevant galaxies based on random parameters. Because deep learning requires a large number of images and parameter samples, the cost of manual annotation of images is too high; and the expertise of the person conducting the annotation is required. Generating datasets with parameters is a more convenient experimental method.
  • When generating galaxy data with GalSim software, the dimension size of the generated data is also set to random because the dimension size of the real galaxy data varies. The python library is used to generate 50,000 galaxy information, which is partitioned into a training set, a test set, and a validation set in the ratio of 8:1:1. In order to handle galaxy images of different sizes, this paper uses two-dimensional data with a standard size of 128*128 size, which will be different in shape from 128*128 size, cut the center of galaxy images larger than 128*128 pixels, and fill the edges of images smaller than 128*128 pixels.

Second NN

  • The second task fits the data to multiple parameters.
jumengting

Result

Generate data First Method

  • The above network fits the galaxy data to one parameter using a 2D convolutional neural network. The following are the training results for training a single parameter - star magnitude.
jumengting

Generate data Second Method

  • Galsim result.

Ground-Truth data Second Method

  • CANDELS

Else

Waiting...

取自“http://cluster.shao.ac.cn/wiki/index.php?title=Deep_learning_on_galaxy_morphology_profile&oldid=4998”