Jupyter Notebooks for Data Driven Astronomy

Introduction

With services such as Google Colab, Binder, and Gradient, now Jupyter notebooks can be hosted and shared online. Users can run shared notebooks without having to install required packages locally on their machines. Although these services come with many pre-installed packages, some newer packages require manual installation routines.

This repository contains codes for the following:

• H alpha Map, WHAN Map, BPT Map and Metallcity Map overplotted with SDSS g band contours
• Fast.ai CNN Classifier using Galaxy Zoo Kaggle dataset

Notebooks

• Marvin : MaNGA_Data_Analysis.ipynb
• Fast.ai : Bar_Classifier.ipynb

Usage

MaNGA_Data_Analysis.ipynb

This notebook can be used to generate different Marvin maps with optical contour overplots using Plateifu of MaNGA sources. An example of maps is given below

Bar_Classifier.ipynb

This notebook contains code for creating a Convolution Neural Network based image classifier for classifying galaxies into Barred or Unbarred using fast.ai. The notebook uses Kaggle Galaxy Zoo dataset for training.

Barred

Unbarred

The example uses the CNN model with ResNet50 architecture with pre-trained weights. Following data augmentation methods are used:

• Rotation : 0 to 360
• Flip : Horizontal and vertical
• Contrast : 0.75 to 1.35
• Zoom : 0.8 to 1.2

Data Augmentation*

Training

Confusion Matrix

Conclusion/Disclaimer

If you have any questions or suggestions for improvements to this repo, please contact the owners of the repository.

This is not an official Google product.

References

• SDSS-Marvin
• MaNGA
• fast.ai