2
$\begingroup$

I'm a Machine Learning Student and I'm working on Strong Lensing Image Classification. this is a hot topic nowadays but strangely after days of searching and communicating with different people I still can't find an accessible Dataset of these objects!

Do you know any place I can find a simple(or not) Dataset containing images, labels and some extended features of Strong Gravitational Lensings?

$\endgroup$
  • $\begingroup$ I doubt what you are asking for exists. Could you clarify what you mean by "labels"? Also, are you asking for images with extended SL features, or for images and a separate set of coordinates that describe where to find the extended features? $\endgroup$ – Alex Nov 15 '17 at 19:37
  • $\begingroup$ Thanks for your attention @Alex . in the simplest way, i hope to find some dataset with some images and 0/1 labels which say is there a GL in the picture or not and maybe some other related features to GLs like the curvature and Z or like these. there are very limited numbers of papers on this topic and they did use some datasets but i cannot find them! i emailed and till now, no answer either. $\endgroup$ – Babakslt Nov 17 '17 at 18:39
  • $\begingroup$ I've attempted an answer. Could you edit the clarifications from the comment into the question? That makes it easier for others attempting an answer to know what you're looking for. Also, what do you mean by curvature? $\endgroup$ – Alex Nov 17 '17 at 22:38
2
$\begingroup$

The most comprehensive catalogue of strong gravitational lenses that I am aware of is the CASTLeS database with about 120 lens systems. Several of them have extended features, arcs, etc., while others are point sources. They provide the redshifts of lens and source where available, as well as time delays between multiple images and the velocity dispersion (a proxy for the mass) of the lens for a few.

You would have to create a comparison dataset without strong lensing yourself, but that should not be difficult: Go to Hubble's Ultra Deep Field web page, download the images with the same filters as used in CASTLeS and cut out likely looking galaxies. The UDF contains mostly very high-redshift galaxies, so the chance of accidentally picking a strong-lens system is very slim.

(Disclaimer: I've only ever used ACS UDF images, not the NICMOS ones; if it's too hard, just skip the IR data.)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.