These forums have been archived and are now read-only.

The new forums are live and can be found at https://forums.eveonline.com/

EVE General Discussion

 
  • Topic is locked indefinitely.
 

Project Discovery and Button Mashers

First post
Author
Previous Topic Next Topic
Tau Cabalander
Retirement Retreat
Working Stiffs
#41 - 2016-03-10 08:33:59 UTC  |  Edited by: Tau Cabalander
Why I'm done with Project Discovery:

1. I spend a _lot_ of time making a very careful and thorough analysis.
2. The results show I picked the same as most others have picked.
3. I get an accuracy reduction.

I can't advance. I hover around 50% (high 54%). No reward in that for me.


EDIT: And then there are the ones that I know others have missed examining the entire sample, and sure enough I get a reduction.

Suggestion: The sample images need filters too.
Barrogh Habalu
Imperial Shipment
Amarr Empire
#42 - 2016-03-10 12:02:10 UTC  |  Edited by: Barrogh Habalu
Pryce Caesar wrote:
There are cytoplasms that are outside of the nucleus.

Cytoplasm is more of a medium than some sort of organelles.
I may be wrong, but plural there in the sentence bothers me a bit.

After trying a few samples yesterday I realized that I have no bloody idea what I'm doing. I think curator of the project said on Reddit that they were preparing videos with more detailed explanation of what we are looking at and what different structures actually are. That would help. There's definitely more material on that than tutorial presents.

Although I'm getting afraid that this project will be deemed a failure.

Future of T3 cruisers - multi-tool they aspired to be instead of sledgehammer they have become
Sheeth Athonille
Rabid Dogz Mining
#43 - 2016-03-10 12:40:24 UTC  |  Edited by: Sheeth Athonille
My experience so far has been mostly positive. I do occasionally miss stupid ones, but for the most part the consensus seems to be fairly accurate (if missing a few). So far my accuracy rate is 68% and slowly rising, so I don't think it has to do with people spamming that is making accuracy drop.

That said, I do think the rewards need re-worked. A high initial time and accuracy investment before it becomes reasonable would make sense to help keep people from just button mashing, but as it stands it isn't really "worth" doing. Go to any cell lab in college and they'll be more than happy to have you do the same thing lol

Also, if you don't know what you're looking at, please do exit out and try an easier one. It prevents the answer pool from getting diluted by people who don't know so they just select one at random. That's not cheating, it's admitting your limitations P
HPA Illuminator
H P A
C C P Alliance
#44 - 2016-03-10 13:27:40 UTC  |  Edited by: HPA Illuminator
Silvenin wrote:
Let's look at an example... (yes, click "example")

On the left side we have the image itself, at the top with all 3 colors visible, at the bottom with blue filtered out.
On the right side we have my answer (nucleoplasma) and the already accepted answer (nuclear specles).

Either I'm terribly wrong... or the majority of voters are.
If I'm wrong I would very much like to hear why.

And it's not the only one... I take 2 to 5 minutes on each picture, checking and rechecking the possible options and refining my choice trying to be certain.... and as a result, my accuracy rating dropped to 47.3%... and it''s still dropping.

Also the payout is laughable. Mining I get500k isk every 3 minutes semi afk. Here I get 50k for burning a lot of attention in the same timespan. As it stands, it's simply not worth the bother. The only incentive to join project discovery at this point, is wanting to help with actual real world research.

What I would like to see is a thread (or even a subforum section) where truly interested players can post pics and ask for clarification to better understand the differences and be able to give better results in the future. And answers should be given only by real researchers (who are behind this), or community members who are marked as knowledgeable by said researchers. but I guess that's just a dream.


As one of the researchers from HPA working with data, I would say that the image you posted shows both nucleoplasm and nuclear speckles... And since the nuclear speckles are really prominent, they should absolutely be clicked. Maybe we should be clearer that there often are many patterns co-occuring, approx. 60% (if I remember correctly) would show more than one location. 2-3 is common, whereas 4-5 is really uncommon.

As for the thread or subforum, there are 4 ppl from HPA signed up here to help out with questions (and since I don't have any of programming skills that HPA_Dichroic does & therefore will just ignore questions related to that, I'd be more than happy to discuss difficult images instead!). However, I'm a complete noob at EVE & this forum, but if you start something like that, I'd absolutely hang out there (hehe, Dr. Lundberg actually asked us to do this during working hours :))
HPA Illuminator
H P A
C C P Alliance
#45 - 2016-03-10 13:30:22 UTC
Saint Athanasius Trigentia wrote:
OK. I'm thinking there are lots of people playing this game that don't bother looking at all the example pics.

In other news, I'd like to see a sub board put up for this or at least a sticky thread to share thoughts on crazy difficult ones.


Please start one! I would be happy to discuss such cases!
Troy Cintryx
Strategic Operations Inc.
#46 - 2016-03-10 13:59:55 UTC
I think "cell to cell variations" should be relatively obvious when you first look at the specimen.

Everything else after that requires a little more analysis.


I think a lot of people really are trying to get this right (This is EVE... I would argue we are the most intelligent MMORPG playerbase in the world), but are having limited success and getting frustrated.

Building empires in EVE for over 12 years.

Watch my EVE Corporation Management training videos here.
Sheeth Athonille
Rabid Dogz Mining
#47 - 2016-03-10 14:09:09 UTC
HPA Illuminator wrote:
Silvenin wrote:
Let's look at an example... (yes, click "example")


As one of the researchers from HPA working with data, I would say that the image you posted shows both nucleoplasm and nuclear speckles... And since the nuclear speckles are really prominent, they should absolutely be clicked. Maybe we should be clearer that there often are many patterns co-occuring, approx. 60% (if I remember correctly) would show more than one location. 2-3 is common, whereas 4-5 is really uncommon.

As for the thread or subforum, there are 4 ppl from HPA signed up here to help out with questions (and since I don't have any of programming skills that HPA_Dichroic does & therefore will just ignore questions related to that, I'd be more than happy to discuss difficult images instead!). However, I'm a complete noob at EVE & this forum, but if you start something like that, I'd absolutely hang out there (hehe, Dr. Lundberg actually asked us to do this during working hours :))


I think that's one of the problems with the tutorial is it only allows for one choice. So I imagine it tends to cause people to focus only on whats most obvious, and not some of the secondary features.

Question, when should we check the Abnormal Sample button? When a cell looks like it's exploded (nucleus outside the cell), or something else?

Also, when a sample has both cytoplasm and the nucleus marked, the nucleus tends to take on a darker green, or would it all be a uniform color? I've seen it both ways, and I'm assuming it usually takes on a darker color, but wasn't sure.
HPA Darkfield Oramara
H P A
C C P Alliance
#48 - 2016-03-10 14:29:24 UTC
Barrogh Habalu wrote:
Pryce Caesar wrote:
There are cytoplasms that are outside of the nucleus.

Cytoplasm is more of a medium than some sort of organelles.
I may be wrong, but plural there in the sentence bothers me a bit.

After trying a few samples yesterday I realized that I have no bloody idea what I'm doing. I think curator of the project said on Reddit that they were preparing videos with more detailed explanation of what we are looking at and what different structures actually are. That would help. There's definitely more material on that than tutorial presents.

Although I'm getting afraid that this project will be deemed a failure.


I'm one of the researchers for this project and if you haven't seen it already there is a tutorial video up know created by my colleague HPA_Dichronic. https://www.youtube.com/watch?v=TrqCWg0cZSk&feature=youtu.be

Hope that will help you to get started at least! And I know it will be hard in the beginning before you get a feeling for it, but the hope from our side is that people will learn and do better and better. And from what we have seen from the test server, you are doing really good!

And regarding the cytoplasm, that is correct. The cytoplasm is more or less just everything in the cell that isn't nucleus or membrane. Most of the organelles are hence in the cytoplasm, but we are still able to distinguish for example a mitochondrial staining from a general cytoplasmic one.
Memphis Baas
#49 - 2016-03-10 14:44:33 UTC
My first question is: what should we do if we have a pattern that's easily recognizable, but also all these dots spread out throughout the cell, but dimmer and not super-concentrated? Do we click the pattern AND the closest thing to the dots, or do we ignore the dots? Especially confused if the "closest match to the dots" says "these dots should NEVER appear in the nucleus" but I see them in the nucleus.

My second question is: what should we do if the sample is blurry and we can't visually distinguish between short filaments and groups of dots?
HPA Illuminator
H P A
C C P Alliance
#50 - 2016-03-10 14:51:13 UTC
Memphis Baas wrote:
My first question is: what should we do if we have a pattern that's easily recognizable, but also all these dots spread out throughout the cell, but dimmer and not super-concentrated? Do we click the pattern AND the closest thing to the dots, or do we ignore the dots? Especially confused if the "closest match to the dots" says "these dots should NEVER appear in the nucleus" but I see them in the nucleus.

My second question is: what should we do if the sample is blurry and we can't visually distinguish between short filaments and groups of dots?


If it's possible for you to take some screen dumps of what you mean and post in the other thread it would be a lot easier for me to give you accurate feedback:
https://forums.eveonline.com/default.aspx?g=posts&t=472862&find=unread

For the pattern that you see in the nucleus too, are you sure it's not below/over the nucleus (if you see some red microtubes overlapping with the blue, then you know that there's some bleed through from outside the nucleus)?
HPA Darkfield Oramara
H P A
C C P Alliance
#51 - 2016-03-10 14:55:37 UTC
Jack 'GUN' Morgan wrote:
A lot of people are using Cytoplasm (first option) as their go to! I've come across way loads of samples where this has a higher a percentage to the correct answer.

Note that the description for this clearly states, 'Is seen throughout the whole cell, except in the nucleus (blue marker)." But on the samples I've seen that people have chosen this for have CLEARLY had green dots all inside the nucleus (blue marker).

I'm loosing the point of doing this as the results we as a community are giving is garbage!


What me and my colleagues in the research team have seen so far is that you are not doing bad at all! The result from the test server was actually really good. Regarding the description of the cytoplasmic staining, there sure should be no staining in the blue area if the protein is localized to the cytoplasm only. However, there's very common that the protein is located to two organelles or more, which means that a cell with both cytoplasmic and nuclear staining will have staining in both the area overlapping with the red and the blue marker. This is a bit tricky of course, but usually they don't have the exact same intensity which might make it possible to differentiate them.
Jack 'GUN' Morgan
Caldari Provisions
Caldari State
#52 - 2016-03-10 14:58:41 UTC
HPA Darkfield Oramara wrote:
Barrogh Habalu wrote:
[quote=Pryce Caesar]There are cytoplasms that are outside of the nucleus.


I'm one of the researchers for this project and if you haven't seen it already there is a tutorial video up know created by my colleague HPA_Dichronic. https://www.youtube.com/watch?v=TrqCWg0cZSk&feature=youtu.be



This is great thank you, but why oh why was this posted to Reddit before the OFFICIAL Eve forums? Roll
HPA Illuminator
H P A
C C P Alliance
#53 - 2016-03-10 15:03:29 UTC
Sheeth Athonille wrote:
HPA Illuminator wrote:
Silvenin wrote:
Let's look at an example... (yes, click "example")


As one of the researchers from HPA working with data, I would say that the image you posted shows both nucleoplasm and nuclear speckles... And since the nuclear speckles are really prominent, they should absolutely be clicked. Maybe we should be clearer that there often are many patterns co-occuring, approx. 60% (if I remember correctly) would show more than one location. 2-3 is common, whereas 4-5 is really uncommon.

As for the thread or subforum, there are 4 ppl from HPA signed up here to help out with questions (and since I don't have any of programming skills that HPA_Dichroic does & therefore will just ignore questions related to that, I'd be more than happy to discuss difficult images instead!). However, I'm a complete noob at EVE & this forum, but if you start something like that, I'd absolutely hang out there (hehe, Dr. Lundberg actually asked us to do this during working hours :))


I think that's one of the problems with the tutorial is it only allows for one choice. So I imagine it tends to cause people to focus only on whats most obvious, and not some of the secondary features.

Question, when should we check the Abnormal Sample button? When a cell looks like it's exploded (nucleus outside the cell), or something else?

Also, when a sample has both cytoplasm and the nucleus marked, the nucleus tends to take on a darker green, or would it all be a uniform color? I've seen it both ways, and I'm assuming it usually takes on a darker color, but wasn't sure.


Good questions! I'll forward the input regarding the tutorial.

Reg. Abnormal it should be used to highlight patterns in the green staining that look specific but don't fit any of the categories. We recently added "rods and rings" after having flagged images internally that had this never seen before pattern.

It's more common that the nucleus has a stronger color than the cytoplasm but that is partly due to bias when acquiring the images. The cell is kinda like a fried egg, and we image a thin slice of it. Due to the nucleus being really big, more of it will appear in the image. Biologically speaking, one would expect to see some images with stronger green in the nucleus and some with stronger green in the cytoplasm,depending on how much of the protein is in each compartment. As long as you can distinguish between the two, choose both. If the color is really smeared out, it should be labeled as not identifiable (not sure what term is used in the game).
HPA Darkfield Oramara
H P A
C C P Alliance
#54 - 2016-03-10 15:04:21 UTC
Nana Skalski wrote:
Gliese Casserres wrote:
Memphis Baas wrote:
Demica Diaz wrote:
I had feeling that something like this will happen. Lol


Me too.
I'm all up for griefing, corp theft, scamming, awoxing, any other form to ruin someone's game, but (to attempt) to ruin something with real world importance makes me sick.

Luckily a simple solution exists. A timer. Nobody cracks the image under 10 seconds.

I sometimes dont know what to choose, because I can identify the structures, but they are not on the list anywhere, like these specles autside cells. They looked like some grains spread around the red structures, like you would drop some sugar grains around it. I have pondered about that image longer than 10 seconds.


Can you post an example of such staining? I think what you describe could be focal adhesions. Do they look simliar to this? http://www.proteinatlas.org/images/4835/78_B12_2_blue_red_green.jpg
Duke Killem
Doomheim
#55 - 2016-03-10 15:09:21 UTC
So what happens when a large alliance decides to tell all it's members to use PD and ONLY choose cytoplasm for all results? I presume they get to the goodies faster.

In a game where people screw over each other is core gameplay for some PD doesn't really have the right audience!
Krevnos
Back Door Burglars
#56 - 2016-03-10 15:30:30 UTC  |  Edited by: Krevnos
Jack 'GUN' Morgan wrote:
Jack 'GUN' Morgan wrote:
A lot of people are using Cytoplasm (first option) as their go to! I've come across way loads of samples where this has a higher a percentage to the correct answer.

Note that the description for this clearly states, 'Is seen throughout the whole cell, except in the nucleus (blue marker)." But on the samples I've seen that people have chosen this for have CLEARLY had green dots all inside the nucleus (blue marker).

I'm loosing the point of doing this as the results we as a community are giving is garbage!


http://i.imgur.com/13qnV3q.png

Case and point!


Actually I can clearly see speculation in the cytoplasm there. Indeed there is also speculation either overlying or within the nucleus.

It is very important to learn how to filter your channels appropriately. Try turning off the red channel and inspect again. Now try flicking the blue channel on and off to see how distribution changes inside the nuclear envelope. Remember cytoplasm also overlies the nucleus, as does plasma membrane, though typically cytoplasmic 'depth' will be lower where the nucleus lies. If the protein appears fainter in the nucleus with no obvious foci, then you can regard it as likely absent from nucleus.
HPA Darkfield Oramara
H P A
C C P Alliance
#57 - 2016-03-10 15:35:47 UTC
SurrenderMonkey wrote:
Haven't played with it much yet. I'm kind of wondering about that what-looks-to-be-an-image-ID in the lower right hand corner.

If that's what it is, I expect we'll have a database of known-good answers and bots to supply them, before long.:\

Also it seems super easy to cheat by just closing the window if you get a hard one. I've seen a few where I just went Shocked at the cluster-F of green, closed it, and opened it back up to a nice, clean Nucleoplasm.


The images are re-named before they enter the game so there's no risk of finding the IDs anywhere else!

Of course that could be a possibility, but after a while when more and more images reach a consensus answer, and hence not be available in the game anymore, there will only be the hard ones left Blink But on the other hand, the whole objective of the project is to help us with the hard ones that we are not sure about, so just give them a try even though you're not perfectly sure!
Nana Skalski
Taisaanat Kotei
EDENCOM DEFENSIVE INITIATIVE
#58 - 2016-03-10 15:40:58 UTC  |  Edited by: Nana Skalski
HPA Darkfield Oramara wrote:
Nana Skalski wrote:
Gliese Casserres wrote:
Memphis Baas wrote:
Demica Diaz wrote:
I had feeling that something like this will happen. Lol


Me too.
I'm all up for griefing, corp theft, scamming, awoxing, any other form to ruin someone's game, but (to attempt) to ruin something with real world importance makes me sick.

Luckily a simple solution exists. A timer. Nobody cracks the image under 10 seconds.

I sometimes dont know what to choose, because I can identify the structures, but they are not on the list anywhere, like these specles autside cells. They looked like some grains spread around the red structures, like you would drop some sugar grains around it. I have pondered about that image longer than 10 seconds.


Can you post an example of such staining? I think what you describe could be focal adhesions. Do they look simliar to this? http://www.proteinatlas.org/images/4835/78_B12_2_blue_red_green.jpg

They were round, not elongated, clear round points of green. they were not attached to tthe red in such way. Rather like you would have tiny round bacteria garnering to the red walls around the cell. And there were also green colored speckles inside the blue. I have seen only one such sample. I will post image here If I will see it again.

ᴇᴠᴇʀʏ ᴘᴀʀᴛ ᴏғ ᴀ ɢᴀᴍᴇ ʜᴇʟᴘs ᴛᴏ ᴛᴇʟʟ ᴀ sᴛᴏʀʏ 📕

ᴡʜᴇʀᴇ ɪs ᴀɴɢʀʏ ᴄᴏɴᴄᴏʀᴅ ɢᴜʏ ᴡʜᴇɴ ʏᴏᴜ ɴᴇᴇᴅ ʜɪᴍ

ᴏsᴘʀᴇʏ 🚀

≡⋁≡ ♥
HPA Darkfield Oramara
H P A
C C P Alliance
#59 - 2016-03-10 15:44:17 UTC
Duke Killem wrote:
So what happens when a large alliance decides to tell all it's members to use PD and ONLY choose cytoplasm for all results? I presume they get to the goodies faster.

In a game where people screw over each other is core gameplay for some PD doesn't really have the right audience!


Since we have some images that we already know the answer to, we will filter you out if you get too many of these totally wrong. That also applies if you select the same location over and over.
Duke Killem
Doomheim
#60 - 2016-03-10 15:51:55 UTC
HPA Darkfield Oramara wrote:
Duke Killem wrote:
So what happens when a large alliance decides to tell all it's members to use PD and ONLY choose cytoplasm for all results? I presume they get to the goodies faster.

In a game where people screw over each other is core gameplay for some PD doesn't really have the right audience!


Since we have some images that we already know the answer to, we will filter you out if you get too many of these totally wrong. That also applies if you select the same location over and over.


Phew....that restores a little faith that this isn't happening.

Cytoplasm, nucleus, cytoplasm, nucleus, cytoplasm, nucleus, cytoplasm, nucleus, GONG, GONG, GONG
Cytoplasm, nucleus, cytoplasm, nucleus, cytoplasm, nucleus, cytoplasm, nucleus, GONG. GONG. GONG