7/21/2014 Update from AsteroidZoo

• by Dr.Asteroid scientist, admin

  Update 7/21/2014

  Welcome to the update for AsteroidZoo!

  I’m deeply impressed by how people are working to find asteroids. Thank you and wow. I also want to give you a feel for how things are going on the backend.

  Images

  So far, you’ve looked at ~30,000 sets of images. This number is incredible, and we are confident that within the imagery you have looked through – asteroids are being found. Nice work.
  We have heard from many of you about an issue with some data sets and not being able to see the imagery, or the images render in a way that makes it hard for you to detect objects. We appreciate the feedback, and are actively updating our image creation software to ensure these issues are alleviated so you can find as many asteroids as possible!

  Detections

  As of this morning, you’ve clicked 761,800 times within these 30k images to identify asteroids and artifacts. While I can’t confirm how many asteroids have been discovered yet, I can say that there are appear to be quite a few discovered that aren’t in the records that I have access to. If I seem slow to answer directly, folks who spend a lot of time with scientists know that we rarely share results until we are nearly certain.

  Here’s an update on several topics that I’ve seen discussed and I want to provide a sticky to have the conversation.

  Science Motivation

  There are several parts to what science we want to get out of this project.

  First and foremost: find more asteroids. The more we know about asteroids, the better. While Catalina does an excellent job on discovering them, some are surely missed because humans are often better at drawing connections between low signal to noise events that a machine can’t – particularly when an asteroid just barely misses a star or is moving through a crowded region.

  Second: a huge advantage of the data gathered on this project will be our ability to dig through the old records and discover asteroids that were known to exist, but we didn’t have a lot of data about the exact orbit. Most of the time, you need to see an asteroid at least twice at different apparitions to confirm the orbit where the asteroid gets its number (i.e., 101955 Bennu, which was 1999 RQ36 until it was seen twice). The amazing work you are doing by identifying asteroids with Asteroid Zoo will pint point more accurate data about asteroids that we have only seen once. YOU will discover them on their second orbit, allowing us to identify crucial and accurate orbital data. Sometimes you may identify a asteroid farther back than it’s first observation (called precovery), which allows us to understand the asteroid orbit far into the future.

  Third: this work will also serve to determine how well the Catalina Sky Survey is finding asteroids, by having detection data to measure against. Do they find 90%? 70%? We are not sure. But once we understand what’s been missed in the images, improvements can be made to the algorithms that Catalina uses to find asteroids. This data will even be used in machine learning, so we can try and find all asteroids (even faint ones the naked eye would miss) in the imaging data through algorithms.

  Public Data

  A few people have also asked about the data itself being made available. The answer is yes! The dataset we’re using will be made fully public. We’re working in partnership with Amazon Public Datasets to make this happen. Once we have progressed further in the Asteroid Zoo detection timeline, we will focus on curating the front end of this dataset. We will update you on the timing of this, once we know more.

  Finally, please send us all your feedback or request about Asteroid Zoo! . We are compiling a list of all requests, and while we cannot fulfill them all, we will do our best to continue to make Asteroid Zoo as effective as possible!

  You’re doing amazing work and I am eagerly awaiting the results of our first asteroids.

  Cheers and good hunting!

  Matt Beasley (Dr. Asteroid)

  Posted July 21, 2014 11:24 PM

• by meegja

  Thanks for this insight!

  Posted July 22, 2014 5:34 AM

• by wassock

  Hi there, what is the size of the total potential data set, 30k images out of how many?

  Posted July 22, 2014 12:58 PM

• by Dr.Asteroid scientist, admin

  We have 997,000 images currently in the catalog (i.e., processed and prepared in the way that AsteroidZoo can use them). But that is roughly 1/8th of the total amount of data we have.

  So, it will be a while before we run out - Catalina has been operating for 11 years in a way that makes it sensible for us to use the data. And continues to take data.

  Some of the currently prepared images will be updated when we improve the image program. (so you don't have to work through a million images before we improve).

  Posted July 22, 2014 3:58 PM

• by stonepenny

  Thank you. I think I've found planets - can this info be used?
  AAZ0001ja4
  #asteroid or planet? Above biggest middle star, going from right to left, tracked over 4 marks (in a semi circle!)

  Posted July 23, 2014 2:34 AM

• by wassock

  Hi stonepenny, think you sre seeing an srtifact from the imaging system, more about it here http://talk.asteroidzoo.org/#/boards/BAZ0000003/discussions/DAZ00003vh

  Posted July 23, 2014 9:30 AM

• by stevef602

  Dr.,

  Thanks for the info. Just one Q: How many times is a set of images reviewed before it's deemed to have it's information identified and categorized ?

  On the suggestion side, an FAQ page. Also, when we are reviewing sets, maybe the ability to select a category for the different types of artifacts you may be interested in quantifying.

  Cheers

  Steve

  Posted July 23, 2014 5:56 PM

• by Dr.Asteroid scientist, admin in response to stevef602's comment.

  Sure - right now we're collecting information on how many people have to see something before it's done, we expect that we're doing a slight overkill in terms of numbers of users looking at it (it's more than 20). When we have a good understanding from a statistical base we'll decrease the number to where it needs to be.

  We thought about artifact designations - one of the end goals is machine learning - and from there the ability to ignore something appears to be more important than specifically knowing what it was. So, in order to not have many additional buttons (star bleed, dead row, cosmic ray, weird thing...) we elected to keep the interface simple. We're looking for asteroids rather than defects in astronomical images - if that helps understand our motivation.

  Thank you for your attention!
  Cheers

  Posted July 23, 2014 8:41 PM

• by calin_uio

  Thanks for the very interesting update, so there are around 3%+ processed images in the first set out of 8. The data scientists here are eagerly waiting for the data-release, and if you somehow manage to keep the commercial implications out of the project - asteroids worth trillions kind of headlines - then there is a solid future for asteroidzoo within the other citizen science constellation. The project has now an incredible momentum, and it's great to be a part of it!

  Posted July 24, 2014 6:23 PM

• by a349

  Thanks for the update, please continue.
  Please concentrate more efforts on the unknown moving object that we discover.
  Please add another General Board for results.

  Posted July 25, 2014 7:22 AM

• by Barbalbero

  It is nice to read this update about the contribution we gave up to now. I hope there will be soon info also about the asteroids discovered by people participating to this project in order to quantify our contriution.

  Posted July 25, 2014 8:08 AM

• by stonepenny

  Hi wassock - what I saw went in a semi circle around a star - not the artefact pattern described in the link (reversed '4'). ie 1 was west, 2 was north west, 3 was north east, 4 was east... all 4 made an arc (not a square or reversed '4'). I've seen this several times now, in 3 - 4 frames around a star. Thanks anyway.

  Posted July 28, 2014 11:05 PM

• by Barbalbero

  Hi to all
  I just want to ask a couple of information.
  It sometimes happens that we observe in the frames asteroids already known. If I remember well, for each asteroid already known there is also the indication of the day of the last observation. If the sets of images we use to "see" the already known asteroids have been taken after the last observations day, I think there can be a kind of update about the last observation. In that case, is it possible to have a kind of "credit" for having done the most recent observation?
  Moreover, in the first post of Dr. Asteroid, there is written: "a huge advantage of the data gathered on this project will be our ability to dig through the old records and discover asteroids that were known to exist, but we didn’t have a lot of data about the exact orbit. Most of the time, you need to see an asteroid at least twice at different apparitions to confirm the orbit where the asteroid gets its number (i.e., 101955 Bennu, which was 1999 RQ36 until it was seen twice). The amazing work you are doing by identifying asteroids with Asteroid Zoo will pint point more accurate data about asteroids that we have only seen once. YOU will discover them on their second orbit, allowing us to identify crucial and accurate orbital data. Sometimes you may identify a asteroid farther back than it’s first observation (called precovery), which allows us to understand the asteroid orbit far into the future". Will be credited for these kind of observations too?
  Thanks for everything and I hope to have good news about the asteroids discovered with this project.

  Posted August 2, 2014 10:30 AM

• by Dr.Asteroid scientist, admin in response to Barbalbero's comment.

  To answer your question: if we contribute to the orbit determination of an asteroid, we'll get notified that our observation was accepted. There will be a permanent record of the contribution on the Minor Planet Center.

  If we catch something that has never been seen, we'll get credit for that.

  Hope that helps,

  Posted August 5, 2014 8:48 PM

• by Barbalbero in response to Dr.Asteroid's comment.

  Thanks for the information!

  Posted August 6, 2014 8:11 AM

• by Barbalbero

  Are there other information about when we will know some scientific results about the work we did?

  Posted August 15, 2014 9:33 AM

• by Meanjean4321

  i would love to see as many examples of asteroids as possible. i mark some that i am sure that are but someone else says they have to show in all 4 frames.

  Posted August 26, 2014 1:16 AM

• by dudrea

  4 frame is indication of good probability But sometimes 3 could be good (asteroid exiting the field, or faint one)
  I allways consider 2 frames observations is too loose. Don't even speak of one !!

  Posted August 26, 2014 11:40 AM