PDA

View Full Version : Should we try to mark on bad images?



Nasiar99
2012-May-04, 07:38 AM
I've been seeing alot of poor quality images with bad subtraction and really bright stars but some of them you can clearly see a blot standing out thats different than all the others. Is it worth marking or just mark them bad image?

example below

http://i1264.photobucket.com/albums/jj491/nasiar99iceinvestigaters/m1105_MB01_27_4-s_0368_0000_A.png

econseil
2012-May-04, 11:32 AM
Hi,

As the substraction process gives an awful image, with squared shapes appearing everywhere, I wouldn't mark anything on such an image. If a process can draw such things, it can also add signal where there's none or delete signal where there's some.
Being different is not enough to be a transient.
There are examples of what a transient look like, and this one doesn't match.

Nasiar99
2012-May-13, 10:34 PM
Well here another example of poor subtraction I have been coming across alot lately.

At the bottom of the image there are many spots people would mark if they where just looking at the examples. Nice solid white blobs.
http://i1251.photobucket.com/albums/hh551/nasiar99/m1105_MB03_04_2-s_1840_0000_A.png

tensegrity
2012-May-14, 04:13 PM
@Nasiar99 Over at IceHunters (my user name was ylla) I saw this type of image frequently. I ignored white blobs at the edge of an image when the line of demarcation between the subtracted area and the unsubtracted area (where the blobs appear) was clearly visible, as shown in your example image. Warm greetings to my previous and new co-investigators.

elelectronlibre
2012-May-14, 04:37 PM
I did (and now do here) the same than tensegrity. I ignored that part of the image, and mark it as "bad image". But if I see a blob in another side of the picture, I mark it.

DanishDynamite
2012-May-15, 09:14 PM
Should this be marked as a bad image?:

http://www.cadc.hia.nrc.gc.ca/data/pub/vospace/KBOzoo/m1105.d/m1105.MB03.31.1-s.1840.0736.A.png

I assume the CosmoQuest team has a purpose in giving us the possibility of marking bad images. I think we could help the team better if we were given some guidelines and examples of what to mark and what not to mark.

elelectronlibre
2012-May-15, 09:42 PM
I don't know whether that should be marked or not as bad image, but I do.

In Ice Hunters project, there were a lot of features you could mark in a "bad image" (Bright star, etc), and here the project team is the same (I guess, not sure here) so maybe they could do the same...

I think we need better communication with the science/software team.

DanishDynamite
2012-May-15, 10:08 PM
I think we need better communication with the science/software team.

I very much agree!

The reaction time to questions from us, the Contributors doing the work, by the science/software team is not satisfactory. My enthusiam with this project is declining, both due to this and because it seems that there are no new members coming in.

econseil
2012-May-16, 10:27 AM
Should this be marked as a bad image?:

http://www.cadc.hia.nrc.gc.ca/data/pub/vospace/KBOzoo/m1105.d/m1105.MB03.31.1-s.1840.0736.A.png


I won't mark this image as bad. There's a third of the image that can't be used, but the other 2 thirds are great. A bad image is an image from which you can't get any data (very bright star, image like the one Nasiar99 showed, ...)

And I agree on the little blobs at the bottom of his image. It's just a problem of substraction. There's a line delimiting the substraction. All the blobs under that line are stars from the original image. Shouldn't be marked.

StarStryder
2012-May-16, 05:09 PM
Hi all,

I'm going to work on summarizing what your finding on a new Bad Images page. I'll add a link to this thread at the bottom of the page. One of the issues with this project is I see the bad image types for the first time the same time you do... So we're all learning to sort through things together.

I'm going to subscribe to this thread, and your posts will go to my inbox, and I'll pipe up with what I know when I can.

Cheers,
Pamela

Nasiar99
2012-May-17, 02:23 AM
Is this what asteroids look like now or just wierd subtraction? :p

http://i1251.photobucket.com/albums/hh551/nasiar99/m1105_MB04_30_1-s_0368_0000_A.png

StarStryder
2012-May-17, 03:51 AM
I've been seeing alot of poor quality images with bad subtraction and really bright stars but some of them you can clearly see a blot standing out thats different than all the others. Is it worth marking or just mark them bad image?

example below

http://i1264.photobucket.com/albums/jj491/nasiar99iceinvestigaters/m1105_MB01_27_4-s_0368_0000_A.png

To me those look like variable stars. The software had something in both images to try and subrtact, but clearly failed, and that usually is a sign of variability.

Shigeru
2012-May-19, 10:49 PM
Interesting, I was having similar doubts with some images myself.

DanishDynamite
2012-May-20, 12:36 AM
Can I ask what the purpose of marking bad images is? What does the CosmoQuest team do with this information?

StarStryder
2012-May-20, 03:56 AM
Bad Image information is generally used as a flag that any marks someone makes on an image should be double checked. This means that if 15 people see an image, and 4 people mark random spots, but 5 mark it as a bad image, those 4 random spots will be given a low probability of being real. If no one marks an image as bad, and 4 or the 15 people all mark the same spot, it is probably real, and if all 15 mark the same spot .... well that goes to the top of the stack.

DanishDynamite
2012-May-20, 03:50 PM
Bad Image information is generally used as a flag that any marks someone makes on an image should be double checked. This means that if 15 people see an image, and 4 people mark random spots, but 5 mark it as a bad image, those 4 random spots will be given a low probability of being real. If no one marks an image as bad, and 4 or the 15 people all mark the same spot, it is probably real, and if all 15 mark the same spot .... well that goes to the top of the stack.
Ok, thanks for the info. This gives me a better idea of which images to mark as bad. I will thus no longer mark images where a third of the image is black/missing as a bad image. Only images where there are false white blobs due to a glitch, will be marked as bad. Some images though have both glitches and what appears to be real transients, so remember not to through the baby out with the bath water! :)