Asteroid Zoo Talk

Can we get some image sets to write our own software algorithm ?

  • marcvanriet by marcvanriet

    Hello asteroidzoo team,

    There are probably quite some of us participants who are professional or hobby programmers.
    Would it be possible to provide a sets of images so that we can try and write our own software algorithm to find asteroids and anomalies ?

    When looking at the images, I am pretty sure that I could write something that identifies many of the anomalies.
    With these filtered out, there would be less images that require visual inspection.
    Also the user interface could be modified so that anomalies found by this algorithm would just have to be acknowledged by the user, instead of the user having to indicate each of them (which is very cumbersome).

    We're happy to contribute to science by searching through hose images visually.
    Why not also let us contribute to science through our programming skills ?

    Maybe you could provide us with some sets of images that are already manually processed. And also a text file with the findings of those images. Then we could try and develop an algorithm and compare with the manual findings.
    If you are reluctant to do so, maybe you can provide only a small set (50 image sets), and then give more sets privately to developers who find good algorithms.

    Best regards,
    Marc
    (electronics designer with 25 years of programming experience)

    Posted

  • marcvanriet by marcvanriet

    I think I found something : http://community.topcoder.com/longcontest/?module=ViewProblemStatement&rd=15948&pm=13093

    Are you using the same data sets ?

    Regards, Marc

    Posted

  • Dr.Asteroid by Dr.Asteroid scientist, admin

    That topcoder program is using the same datasets. There will be a collection of data available to anyone who registers - even if you don't enter in the timeframe allotted. If you want to develop your own moving object pipeline, this might just be the cash prize you're looking for.

    We (Planetary Resources) are working to put all of this data in a publicly accessible repository with Amazon Public Datasets. But you can imagine we're pretty busy - and that will take a lot of work. There are a lot of issues to make sure we get straight to ensure that the data is good and well organized. Not quite enough time/energy to make sure the details are set up correctly in the next few months.

    TL;DR. Topcoder will have data soon. Join up. All the data will be available, but I don't know when.

    Posted

  • marcvanriet by marcvanriet

    Just wondering... the data sets of topcoder have images of 64x64 pixels. Those of Asteroidzoo are 256x256.

    Do you enlarge them for viewing, or do you make your own sets out of some larger images ?

    Best regards,
    Marc

    Posted

  • Dr.Asteroid by Dr.Asteroid scientist, admin

    The data in the next topcoder match will be the full 4k x 4k image. The ones you're familiar with with for the first contest that was to determine if a particular object was a real or false detection. For that work, only the thumbnail around the object was required.

    Posted