eahussein / dc1

This tutorial is based on the SKA Data Challenge 1. The aim of the tutorial is to learn to identify and classify sources is radio images. The data provided is simulated, to represent what the SKA data will look like once the telescope is in operation.

The aim of the tutorials is as follows:

• Source finding (RA, Dec) to locate the centroids and/or core positions,
• Source property characterization (integrated flux density, possible core fraction, major and minor axis size, major axis position angle)
• Source classification (one of SFG, AGN-steep, AGN-flat)

Data

3 different simulated data are used in this workflow, where the simulation represents the following frequencies:

• 560 MHz
• 1400 MHz
• 9200 MHz

>  bash binder/download_sample_data.sh

Hackathon Task

From the proposed pipeline, investigate new ways to find/classify sources.

Prerequisites

All the libraries/dependencies necessary to run the tutorials are listed in the requirements.txt file.

Installation

All the required libraries can be installed using pip and the requirements.txt file in the repo:

> pip install -r requirements.txt

Would you like to clone this repository? Feel free!

> git clone https://github.com/Hack4Dev/dataChallenge_hack.git

Then make sure you have the right Python libraries for the tutorials.

New to Github?

The easiest way to get all of the lecture and tutorial material is to clone this repository. To do this you need git installed on your laptop. If you're working on Linux you can install git using apt-get (you might need to use sudo):

apt install git

You can then clone the repository by typing:

git clone https://github.com/Hack4Dev/dataChallenge_hack.git

To update your clone if changes are made, use:

cd dataChallenge_hack/
git pull

---

dc1's Issues

The change that happens in the number of sources in the different frames

Why does the number of sources change from a simulated frame to another?
Further, does that mean the frame that produces more sources is better than the others?

Primary Beam correction

Doing source finding before cropping the training area

Why do we have to crop the image first and then perform source finding separately on the training and the whole image?

Can't we just perform source finding on the whole image and then split the data into training and testing??

Source finding

we need to add the explanation of What is source finding

Feedback

We need to write an explanation for the following:

• Machine learning
• #18
• #19
• Score

is the code based on the final DC results? ? ?

Adding a threshold parameter to the primary beam correction method

Can we add a threshold parameter to the primary beam correction method, so that learners can do further experiments?

Possible data leakage

We are doing source finding on the whole, and on a training image (cropped from the whole image). Just want to confirm that the sources for the whole images are not included in the training. This can be a problem for assessing the accuracy of machine learning.

Do we need the first two links in the ReadMe files?