Suggestion for an image with too many features to classify
-
by SEJones
hi, I'm a newbie! I am really enjoying this project. however, I have found that on a few occasions, I get an image with far too many features to classify (maybe >=30 individual features of similar nature). the rules of this project are really suitable for images containing a manageable number of features. but the manual pattern recognition process becomes inefficient for busier images. may I suggest that you add two more options in classification toolbar just for these types of images like:' >90% with fan features' and '>90% with blotch features'. You can add suboptions about the predominant wind direction. this still allows users to identify the 'interesting' features.
Posted
-
by gnarlyexpert in response to SEJones's comment.
May I second the above suggestion. The images absolutely covered with hundreds of tiny fans/blotches are making me loose the will to live! I end up logging out & then on again to avoid them, which is surely defeating the object of the exercise? Otherwise great project: beautiful images.
Posted
-
by nick.coram
Yup, I'll third that! Also guilty of going of to do other things when a page with masses of features turns up.
Posted
-
by SEJones
i didn't want to say in my original post but yes, i am also guilty of skipping a few of these types of images (in case i was passing bad practices to other newbies like me)!
Posted
-
by nick.coram
I'm sure lots of people do. its only human. It must skew the data though!
Posted
-
by delete_def2cad5 in response to SEJones's comment.
Me too so don't feel too bad about it SE. I try to mark/classify as many as I can but have skipped several of these really busy pictures and also ones which don't seem to have anything to classify that I can see in the photo, so just click on finish then next (maybe I should log out instead as well, don't want the photos I skip to be added to the count of those I've classified). I want to do the best I can for this project but it's just overwhelming when you get these busy pics especially if you just have three tools we have. I am trying to mark stuff on as many photos as I can though so I hope I'm still making a valid contribution and being of some help
Posted
-
by vlp810
+1
Posted
-
by mschwamb scientist, translator
Please try to not skip the images. That hurts our statistics because then we won't even know that there were fans in the image. If everyone who looks at the image does that, then we don't know there are fans or blotches at all. Please mark as many as you can on the images At least marking a few that you can identify as fans or blotches would let the science team know there are fans or blotches in the images there. I'm sure this is something the team will be discussing soon and seeing what's the best way to get those images classified and not have a 100 fans to mark.
Thanks,
~Meg
Posted
-
by Chocstar
I would happily do all these busy images with many tens of fans but am still having issues with my huge tools that obscure these tiny fans/blotches.
Are they working on this issue ?Posted
-
by mschwamb scientist, translator in response to Chocstar's comment.
There's apparently an issue when mapping lots of fans that causes a memory leak from the java library we're using to make the drawing tools. There's a long term fix that will be implemented. We'll see what we can do in the short term. Apologies for the inconveniences. For now do the best you can, and hopefully we can sort the issue soon.
~Meg
Posted
-
by Wounded_Knee
If i get an image that Iam never gonna be able to classify (number an orientation(fan widen bug)) , or it grinds to a halt half way through ,i refresh page for a new image, as the page hasnt been finished an sent back i presume ( an hope ) this doesnt skew the data set?.
Posted
-
by delete_def2cad5 in response to mschwamb's comment.
I usually do end up marking some things on these photos which have loads of features rather than leaving them completely untouched, but if they have just too many features, I don't mark them all and have to go off to have a break. I don't think I've completely skipped a busy pic without doing at least one or two. There are some I get which don't seem to have anything on them to mark so skip those because I don't really know what I'm meant to do to a pic with no blotches, fans, spiders or much else, don't want to hurt your statistics though. Will try to do better
Posted
-
by SEJones
just to be clear, i did not skip 100% of the really busy images. what i did was to skip a few such images while on most occasions i really made an effort and classified everything beacuse after all i signed up to do this project. but i log out straight away because it took a lot of effort on one image. mschwanm's suggestion to at least do part of the busy image is a good temporarry solution. thanks for reassurance..
Posted
-
by SEJones
by the way thanks for reminding us about the stats issue. the tutorial informs us that the whole project relies on multiple users giving classification data. but if a busy image is consistently skipped over then there will be zero classifcation data for that image (so probability distributions not possible).
Posted
-
by SEJones
may i add another suggestion for software developers to consider? perhaps if you added copy/paste feature to save time?
Posted
-
by gnarlyexpert
Can we get a consensus (even temporary) ? For instance: on heavily featured images, make the average size/orientation of the fan/blotch, but in a regular format, like say four points of a rectangle and then a central point, that viewers could read as shorthand for an image covered with 'like' arranged features. Either all over or covering the size of the above suggested rectangle? Any takers?
-
* *
-
* e.g.
Posted
-
-
by gnarlyexpert
Sorry, I did an example above made from 5 asterisks, but the post window failed to "translate".
Posted
-
by nbardoni
How about a multiple-fan / blotch tool? Could be similar to either of the present tools... only would save loads of time. The new tool could incorporate a text box where users could enter the number of fans / blotches. Would save us going mad. 😃
Posted
-
by AUricle in response to nbardoni's comment.
Where would you 'point' the tool to mark multiples? I think if they had a tool that could mark multiple featues at one time, they wouldn't need us! 😃
Posted
-
by BlueMoon58
Seems to me that defeats the whole purpose of what the folks a JPL are trying to do which is to map the wind speed and seasons on Mars. Just take your time and make a determined effort to record every Blotch & Fan. If the blotches are small just click on each one in the centre. That way you are at least creating a reference point.
Posted
-
by gnarlyexpert
Any progress on tying to resolve the issue being discussed here? ........anyone? ...........hello .....
(no, but seriously, was just wondering!)Posted
-
by mschwamb scientist, translator in response to gnarlyexpert's comment.
If you're referring to the bug that crashes the browser when too many fans are drawn I believe we're working on it. I'll try to inquire what the status is from the development team.
~MegPosted
-
by gnarlyexpert
Apologise for the levity Meg! No, was enquiring about the possible solution to images with too many features to classify: any news?
Posted
-
by amm04yu
Not only memory leaks with the fan shapes but also blotch tool.
Caused my browser to crash and close.
Alan.Posted
-
by mschwamb scientist, translator in response to gnarlyexpert's comment.
The best thing to do right now is mark as many fans or blotches before the browser slows down and crashes that way we'll at least now there are some fans and blotches in the images.
Thanks,
~Meg
Posted
-
by di_twiss
In some cases, I have just enveloped the fans or blotches, within one large 'fan' or 'blotch'. I hope this is acceptable? At least it still registers that there is something there.
Posted
-
by mschwamb scientist, translator
yeah. if you can mark a few individually that helps more because if you marked a few then someone else who saw a few in that same image but will like mark different ones so it gives us more measurements.
Thanks,
~Meg
Posted