Suggestions for improving user experience on AsteroidZoo
-
by
hoekstra
New user here, played with the site for about 2 days. Seems fun, but some things about the site can be a bit frustrating. Allow me to do some simple suggestions, which I believe would improve the user experience a lot (more or less in order of importance):
-
Please align the black/white images as well as possible, they sometimes jump around by tens of pixels! This makes the 'movie' much less helpful than it can be. There is no technical excuse that this cannot be fixed, since the images in the "color-cycle" of the same set are usually aligned by better than 1 pixel.
-
The movie is too slow, please speed it up or make a knob to change the speed. As also shown in the excellent video by a user (http://youtu.be/qtQ7w-d9BhU), the best way to detect movement is to move the slider left and right with the mouse (or with the number keys), since the movie is looping too slow. Sliding all the time with the mouse causes too much strain for the hand.
-
Add more training images of known comets with lower SNR. Adding training images is a good thing, since it keeps you sharp, but most of them are 'too easy' and I assume that they can easily be found by your existing algorithms. You state that your goal is to find more of these hard-to-detect asteroids, so please train the user with harder cases, not only the super easy ones!
-
For images posted in the forum, you can only watch a simple movie in black/white, there is no option to use the color-cycle, invert colors, etc., like in the main classification window. Also in case you missed a known asteroid, there is no way to go back to the standard tools to see why you missed it, so that you can learn from your mistake. Obviously, there should no buttons to submit asteroids/artifacts after you have been told of a known asteroid or in a forum post, but all the other buttons should be there.
-
Streamline the user interface. It costs me 4 mouse-clicks to process an image without any interesting features ('color cycle', 'nothing', 'finished', 'next subject'), this could be brought down to 2 with a button 'I don't see anything interesting, give me the next image'.
-
In forum posts, allow some annotation of the images by different users, like adding an arrow, or simply the cross-marks added during the classification. Now you always have to describe in words where you think you see something.
-
In the user profiles, there are no statistics about how many images you checked, how many asteroids you found, .... Please watch some talks by Luis von Ahn (inventer of Recaptcha, Duolingo) about the gamification of tasks on the internet. If you want to have many happy and 'productive' users, reward them with 'fake internet points' like leaderboards, user scores, statistics, .... anything.
-
In the classification window, there is a button to 'add image to favorites'. In the the summary page, after there you have finished classifying, there is no such button. Also in forum posts, there is no button to favor an image (or I didn't find it).
-
If possible, improve the alignment of the images in the color cycle even more. Sometimes you think you see some typical 'moving rainbow', only to see later that all faint stars move in exactly the same way.
-
Please add a button to report 'bad sets', like ones with missing frames, ugly edges or excessive differences in brightness. Reported bad sets should not be shown again to other users. Better yet, let the computer throw out the bad sets so that we don't have to see them, there is no hope in finding an asteroid in any of these, they are just distracting.
-
Please don't let us mark all the obvious artifacts like hot pixels and flat-field errors. Since these should stay at a fixed location on the CCD, it should be trivial to mark these things with computer algorithms. Again, it is just distraction, and you learn nothing new.
Posted
-
-
by
CTidwell3
Some great suggestion. A couple of quick comments on a few.
For #3. Not sure if you meant images in the training tool when you start or the images that pop up from time to time saying it is a known asteroid. From what I can tell, the later 'training' images are not hand-picked, but instead a process where they run the known location of the image set and the time it was taken against the known database of asteroids. If a known asteroid of a certain magnitude is in that set, it will show the 'There are known asteroids in this image' message along with the green circle where the asteroid is. It is possible that there are other asteroids in those sets as well, and seem to remember a few sets in the past with both known and unknown asteroids in the same image. It can also be good statistical data to know what type of known asteroids can and can't be seen in these sets by users.
For #7. There is a count of images you have submitted, but it's part of the zooniverse functionality so not immediately intuitive that it exists. While on the classification interface pages, if you click on the round globe icon in the top right corner near your name, it will ask you to log in again then take you to a page with a count of submissions for every zooniverse project that you have been involved with, including Asteroid Zoo.
Posted
-
by
Andy_Arg
Good points.
I want to suggest for point #6 a grid, 5 x 5 at least, one that can be activated and deactivated, and make it possible to reference the cells by means of letters and numbers as in a spreadsheet. Say, there is an asteroid in B4 for example.Posted
-
by
dudrea
Oh YES PLUS ONE !!!!!!!!!!!!!!!!!!!!! Cause it's quite a mess now to explain what we see or understand what other people have seen !
Posted
-
by
hoekstra
@CTidwell3
Regarding #3: it would be good to have harder images both in the 'tutorial' and for the known asteroids. I don't know if these last are inserted in the standard queue on purpose (which is not a bad thing) or that they are just appear randomly. I just find it surprising that all the 'known asteroids' are extremely obvious and that I also see many 'this looks like an asteroid, but it is clearly just noise'. I would expect an intermediate category of 'this is probably an asteroid, even though i'm not 100% sure', but I have only seen 1 or 2 of those. We need more training for cases from this last category, because those are the ones a computer might have problems with. They should have some examples of these on file (e.g. because they were confirmed by other images), which they should insert in the queue for training.Regarding #7: Thanks, I didn't know about that, but this information should be available in asteroidzoo itself.
@Andy_Arg Indeed, a simple solution like adding a grid is probably the best.
Some more suggestions:
-
Make all buttons also available as keyboard shortcuts (preferably on the left side of the keyboard, since 1234 already changes frames, and you use the right hand for the mouse). Maybe 'C' for color cycle, 'A' Artifact, 'N' for nothing... Using keys is much faster than moving the mouse up and down the screen and, as a bonus, you make the site accessible for people who might not be able to use a mouse.
-
The menu structure is confusing: The main site and the forums have different menu buttons on the top, which point to different sites (e.g. the profile on the main site shows something else than the profile on the forum). It would be better if these are exactly the same.
Posted
-
-
by
nicro46
I agree with Hoekstra on all points: in particular, paragraphs 4, 5, 8, 9 and 10. This job is interesting but has not the charm of the surveys found in Galaxizoo; in addition a 20% of the images are of BadSet unusable, the remaining to nearly half are technically poor, with excessive background noise that prevents detect faint objects. I asked why so many colleagues continue to bring these unnecessary images in Discussion; Answer: BadSet homogeneous collections can be used to understand and improve.
Mmmm ... There are 6-7 types of defects that are repeated continuously, software bugs but also hardwere, in my opinion, that should be well known.
In addition there are artifacts, which are subject to the foregoing.
Personally, after more than 4000 images examined, I found thirty asteroids known, about twenty possible candidates (of which there is no news if verified !!!! or less), and a good number of asteroids known but which are not was able to detect, some certainly my fault, but most because not visible to default program.
I use first of all the images flicker, if I find something to step sequence in B / W
When the set is of good quality, the stars are fixed points on a black background and if something moves
is immediately visible as three points in a sequence of different color.
But if there is background noise is a disaster.
In conclusion, if the aim is to find the weak thousand asteroids detected by the automatic searches, we need to set better image quality, remove the bad ones,
and information on the outcome of the work done (in order to keep up the interest)Posted
-
by
peterbees
I also agree with Hoekstra and others on almost all points - though some to greater degree than others.
Shortcut keys, better image registration, a marking tool for explaining points of interest in the discussion boards, a 'bad set' button, and streamlined user interface are probably my favourites.As a physical scientist with a background in visual perception (seismic interpreter) I'd also like to add a couple of screen tools to the Wish List:
- A simple, integrated zoom and pan tool (yes, you can use CNTRL +/-, but then you often lose the functionality of the controls)
- An image contrast control. Particularly where there is a bright star in or near the image, the field objects are often washed out. Changing contrast (maybe and brightness) would not totally restore the background, but it would help. A full-set application would be nice, but in some cases a frame-by-frame adjustment capability would help as well - such as where there is a bright transient.
- Capacity to add noise to the data. Sounds wierd, but the color cycler is most useful for faint objects when there is a moderate degree of background noise. It's a great tool, but fairly useless in the high quality image sets (except where movement is so obvious that you don't need it). Some random background noise would enhance the more difficult, faint detail.
And last, please explain to users that there is a long process between seeing a moving object in the photo sets, confirming it as an astonomical object, re-acquiring it to confirm sky-path, calculating orbital parameters, checking for duplication with known or possible corresponding asteroids, and believing that a new asteroid has been located. Let alone convincing others of the fact, and getting object naming panels to sanction a formal name.
Folks, there is NO WAY to confirm a new asteroid sighting in days - it is more like years at best. So you CANNOT get any Attaboys for your work, or a certificate of ownership of an asteroid. Live with it!
Posted
-
by
peterbees
And another thing, or two:
The owners of the project don't have to make many changes for a much better user experience.
For instance, some simple instructions for locating items on an image. I suggest using the clock nomenclature for direction from centre, plus a suffix (1-5, for example) for distance from the centre towards the edge. So, 5:00-2 would be slightly right of centreline, 40% of the way from centre to lower edge.
Also, get the notice boards organised properly (yes, that's also the fault of users, but some direction would help). At present there is really no differentiation between the 3 boards in terms of topics. Give us some clearer headlines and instructions as to what kind of comment/question goes where.
And how come your boards do not accept single carriage returns (or is that just me?)
Personally, I appreciate your efforts in running the project in this format, but you and others in the field have to realise that it will take dedicated effort to maintain a happy and productive work force. (after all, your objective is to find unknown asteroids and such, not to run a social experiment on statistical categorisation of remote computer users, No?)
Somebody has to take the time to respond more often to posters...
Anyway, keep up the good work.
Posted
-
by
nicro46
Following the post of Peterbee and reviewing the comments and requests of fellow volunteers Asteroidzoo, I would like to report my opinion on how to use this program.
All concretely feasible improvements would be welcome, there are motions interesting suggestions about it, I personally do not use the keyboard but only the mouse; search for objects I can only tool Flicker.
I trust the experience of Pererbee says that if a slight background noise can be useful, the problem is the 'excess!
The fact remains that too often after viewing a 'picture where asteroids are not visible, the program warns us with a green circle instead that at least one had to be found! (This is regardless of our ability)It is obvious that the confirmation of a discovery is subject to established a long time, what might
give a little encouragement would know (and what) the images are already promising to 'examination of the scientists, while doubtful were discarded; know how many are
Asteroid candidates and those who have reported
Keep in mind that we have to watch hundreds of volunteers frame (objectively of no aesthetic value and also of poor quality) before finding a possible candidate, without
no gratification.Posted
-
by
CTidwell3
@peterbees - some of the notice board issues are common across all zooniverse projects that use this out-of-the-box board interface. It comes with three categories (Help, Science and Chat), and I too have had issue of carriage returns, particular in the object comments.
The FAQ discussion does give a little direction as to what they thought the three could be used for:
http://talk.asteroidzoo.org/#/boards/BAZ0000001/discussions/DAZ00003w5@nicro46 - I do have that happen to me too, missing a known asteroid, though I often find it is on sets that are over-exposed and lacks enough detail. Sometimes when I go back and look in the area afterward in the object chat window, I can see, but not always. This is not always a bad thing:
http://talk.asteroidzoo.org/#/boards/BAZ0000001/discussions/DAZ000040jI do agree that additional forms of feedback over time will help, since as was mentioned before, this can be a slow process. Myself, I try to look at asteroids that others have commented on in chat to see if I can notice them to. Helps me pick improve finding more myself.
Posted
-
by
hoekstra
I have not visited the asteroidzoo for some time, but its seems that nothing has changed. Any comment from the site moderators about my suggestions?
Posted