The largest known prime number is 17,425,170 digits long and was discovered through distributed computing.

[Project] EyeWire
#1
The human brain has 100 billion neurons, each neuron connected to 10 thousand other neurons. Sitting on your shoulders is the most complicated object in the known universe - Michio Kaku

[Image: 250px-EyeWire-Logo-Blue.png]


Site: http://eyewire.org/

Eyewire is an online game intended to map the brain by analyzing and recreating the 3D images of neurons. The project works similarly to other citizen science projects such as the Zooniverse projects and Foldit, by having multiple users solve the same work units (known in EyeWire as 'Cubes'), while simultaneously using the results to improve the ability for computers to more effectively perform this same task in the future.

As with many citizen science projects that require the user to play an active role in generating results for the project (as opposed to projects like F@H that simply use computing power), EyeWire uses the human brain's ability to effectively process patterns and more subjective things that computers currently cannot do, or are too inefficient to complete.

Volunteers are given one Cube at a time. Each Cube is a representation of a volume a little under 100 micrometers in dimensions, containing a small section of a neuron. The Cube will have a pre-existing part of the neuron within a 3D view, and then a layer-by-layer 2D scan of actual cells beside it. These 2D layers will then have the known section of neuron coloured in. From there, the user must go through the layers and identify which of the cells belongs to the original known part of the neuron.

Upon completion of a Cube, the user will be rewarded points based on accuracy and how much there was to colour in, resulting in a certain score. Each can take anywhere from a few seconds, to a few minutes to complete.

While there is a slight learning curve, understanding the controls and interface is relatively easy. By the time you have completed the tutorials and practice Cubes, you should be able to go through the real ones quite easily.



Below are some tips and frequently asked questions:


How are points calculated?
The number of points earned is proportional to the amount of colouring in that you have done. That is, if you need to colour in a large part of the cube, you will get more points for doing so, compared to something that does not have as much. Your points are then affected by your accuracy, and as such, you will only be rewarded for parts of a cube that have been coloured in correctly.

How is accuracy measured if these are unknown parts of a cell?
Accuracy is determined by comparing your results to several other users to obtain a consensus. While this could potentially mean that an incorrect consensus may be made, moderators known as 'scouts' and 'scythes', as well as admins of the project can identify and correct mistakes, using their experience. Whenever a cube is adjusted by these higher-ranked players, the points and accuracy of any user who worked on that cell will also be adjusted accordingly.

Your accuracy values are then determined using the last 60 cubes you have submitted, assuming you needed to fill anything in, rather than submitting an already good cube. This is important to keep in mind if you like to check your accuracy after submitting each cube. Since this accuracy is an average from your last 60 cubes, your accuracy can rise in two ways: The first is by submitting a cube with an accuracy greater than your current average. The second is when a low accuracy cube becomes older than your last 60, and is no longer part of the last 60 cubes you have completed. The reverse is also true with getting lower accuracy. Keep in mind that L1 and L2 cubes are separate, so having low accuracy in L2 will not affect L1 accuracy.

What is a trailblaze?
Once you get the hang of EyeWire, you will have the opportunity to be the first to work on a particular cube, assuming you have a high accuracy. When you trailblaze a cube, you earn 25 or 50 points, and your results set the standard for which other users will have their points and accuracy measured against. Trailblazes are given randomly, and users will not know if they are trailblazing a cube until after it has been submitted.

Why am I not earning more points from trailblazes?
As you are the first to submit the results of a cube you are trailblazing, there is no data to compare results with, and therefore the server is unable to award you more points at that particular time. Once a few other users complete that same cube over the course of the next few minutes, you will be awarded the same points as if the cube was a normal, non-trailblazed cube, plus the additional 25 or 50 points bonus.

How are other people earning over 1000 points per cube?
Some cells are harder to complete than others. As such, these cells award double the points. Level 1 cubes can earn you a maximum of 1000 points, while Level 2 cubes can earn you up to 2000. To be able to solve Level 2 cubes, you must complete additional tutorials. You can start these by going to the Overview screen, and changing to any Level 2 cell. It is highly recommended that you first get a hang of Level 1 cubes first, or else you may find yourself with an extremely low accuracy in Level 2 cubes.

I can't tell if this part of the cell is supposed to be coloured in!
Sometimes a part of a cell in your cube may be so small, or the cell's walls may be too blurry that it's hard to tell what's what. In almost every case, you can get back on track by changing your 2D plane view. That is, if you were originally viewing your 2D slices in the XY axis, you can change to the YZ or ZX axes. In other words, you can change your 2D views from front view, to top-down, to side view. To change views, simply press the left or right arrow keys on your keyboard. It is highly recommended that you do this even for easier cubes, to ensure you haven't missed any small pieces, especially at the sides of the cubes. Sometimes something might look like it belongs to the cell you're working on, until you change planes and realise that was not the case.

I'm still not sure if I've coloured this cube in right...
A rule of thumb with most cubes is that you should have a piece of a cell that starts from one of the six sides of the 3D cube, and continues until it reaches another side. This doesn't necessarily mean it has to be the opposite side though. That being said, every single branch has to eventually end, so there will be some cubes where the cell won't go to another side. These ends should not be an abrupt or jagged cut-off, so if you come across something like that, flip between the planes and see if viewing from another angle helps. If you're still having trouble, you can always ask in chat for someone to enter your cube and watch you. They can then see what you have done and make any necessary changes for you if need be.

What is a merger?
Mergers are a general term for anything that doesn't belong in the cell that is being worked on. One such example is when a general consensus has been made by users on an incorrect portion of a cell. These mergers should be reported in the chat so that they may be fixed. The AI can also make a merger in the known part of the cell that is coloured in when a cube is loaded. These are called AI seed mergers. Again, they should be reported. In these instances, continue colouring in the rest of the cell, but ignore the merger part. Right-click to un-colour a piece. This may leave you with a broken cell, but this is fine.

My inaccuracy value shows over and under colouring. What is worse?
Over-colouring means you have clicked something that was not part of the cell you are working on, while under-colouring means that you missed a piece that was supposed to be coloured in. Because the goal of the project is to recreate the 3D structures of these cells, it's much harder to find something that is missing, than to remove something that shouldn't be there. If you're unsure whether you should colour something in or not, be adventurous and try colouring it in. Remember that scouts, scythes, and admins are there to catch any incorrect pieces, and they're happy to do it as it earns them points as well, so don't feel nervous about breaking science!
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes: Ruthalas
#2
[Image: Lqa7ijW.png]

First 50K a few days ago.

The sad part is that several players make these sorts of points in a single day.
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#3
Ohohooo yeah, this is cool.  I'm definitely in.

I may suffer some Fallout withdrawals, but a puzzle game that directly connects to neural science is too awesome to pass up!

edit: So how long is the learning curve for this? I've done a few cubes past tutorial and my accuracy is abysmal (under 10% overall). 3 of 5 cubes I've tried have started wrapped around a corner with no apparent jagged edges to continue from? Started skipping the evil corner cubes and my accuracy has greatly improved in two cubes
[Image: sigimage.php?u=617410&t=212997&b=rarity2]
Reply
Likes:
#4
If you have trouble with a particular cube, ask someone in the game chat to watch your cube. Certain members (called mentors) have the power to enter your cube with your permission. They can advise you on what to do, and can even intervene by adding and removing parts themselves. Any parts added by mentors will be shaded in a different colour to distinguish between what is theirs and yours.

As a side note, those cubes with tiny corner pieces are the best for improving your accuracy. If you can avoid skipping them, they will help you out in the long run.

I'm going to update the first post to add a little more information about hoe accuracy is calculated under one of the existing sections...
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#5
It has been really overdue now, but I finally hit 100 000 points, after four months of tracing!

[Image: GrGr3co.png]

Don't get discouraged, guys. I'm just really slow because I use a 10 year old laptop, and I'm frequently out of the country for work!
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#6
(2016-01-13, 05:53:13 PM)hiigaran Wrote: If you have trouble with a particular cube, ask someone in the game chat to watch your cube. Certain members (called mentors) have the power to enter your cube with your permission. They can advise you on what to do, and can even intervene by adding and removing parts themselves. Any parts added by mentors will be shaded in a different colour to distinguish between what is theirs and yours.

That's actually a really cool feature! I had no idea that that was an option, I'll definitely start taking advantage of that if/when I get absolutely stuck.

(2016-01-13, 05:53:13 PM)hiigaran Wrote: As a side note, those cubes with tiny corner pieces are the best for improving your accuracy. If you can avoid skipping them, they will help you out in the long run.

I was afraid you'd say something like that. I've been forcing myself to try and do the tiny corners, but I usually flub up somewhere and only get like 40-60 points. Twilightblush  Ah well. Practice, practice I suppose..


Also, nice job getting to 100k points!
[Image: sigimage.php?u=617410&t=212997&b=rarity2]
Reply
Likes:
#7
You normally wouldn't get more than that many points for corner cubes. It's normal to even get around 5 points. Different cubes have different potential max scores. One cube might be worth only 5 points if you're 100% accurate with it, while another cube could be worth 100, 500, 2000, or anything in between.

The best indicator of performance would be your accuracy rating. Open your profile and hover your mouse over the accuracy bars. They will show how much of your inaccuracy is from colouring in things that shouldn't be coloured in, and not colouring in things that should have been coloured in. over and undercolouring, basically.
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#8
My accuracy seems to be hanging around 93% as well.  So I guess I'm on par.  Sometimes Im afraid to submit my work, feeling like I've missed something major.  But the overall accuracy seems fine.  I also helps to view the 2D image on the three axis before hitting finished.  I scroll through the whole cube quickly, then hit the 'right arrow' key to switch planes, scroll through that one quickly, then hit 'right' again and go through the last one.

[Image: sigimage.php?u=658963&t=212997&b=rainbowdash1]
Reply
Likes: pinormous
#9
I'd have a lot higher score if I could spend more time tracing and less time waiting for everything to load... I miss my old ISP.
[Image: sigimage.php?w=2793449&p=81403693&u=6281...b=octavia1][Image: sigimage.php?e=pinormous&b=vinylscratch1]
Reply
Likes:
#10
(2016-01-19, 11:20:19 PM)pinormous Wrote: I'd have a lot higher score if I could spend more time tracing and less time waiting for everything to load... I miss my old ISP.

I started having that issue yesterday for the first time.  If you are referring to the checkboard issue when scrolling through layers.  I was wondering if it was  on their end or mine.  Frustrating  Flutterrage
Reply
Likes:
#11
From a quick network inspection, it appears each cube downloads roughly 20mb for the entire 256 images on a particular plane. At least, I think it's per plane, since I always seem to be downloading again when the cube switches planes. Assuming it's 20mb per plane, you're looking about about 60mb for the entire cube.

I'd need to double check to confirm though.
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#12
Since a few people are having accuracy issues, I decided to try and create a video to help. This isn't exactly a tutorial, as nothing is explained (no mic). It's more of a playthrough. I tried to slow things down so that people can see what I'm doing though. Best to view it in fullscreen to see the little bits in the 2D slices that are being selected.


As you can see, what starts out as a seemingly innocent branch pair, turns into something that splits a few more times down the track. This is nowhere near the most complex cube you'll encounter, but those sorts of cubes are quite rare anyway. Also, this is a L2 cube.
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#13
(2016-01-24, 07:46:06 AM)hiigaran Wrote: Since a few people are having accuracy issues, I decided to try and create a video to help. This isn't exactly a tutorial, as nothing is explained (no mic). It's more of a playthrough. I tried to slow things down so that people can see what I'm doing though. Best to view it in fullscreen to see the little bits in the 2D slices that are being selected.


As you can see, what starts out as a seemingly innocent branch pair, turns into something that splits a few more times down the track. This is nowhere near the most complex cube you'll encounter, but those sorts of cubes are quite rare anyway. Also, this is a L2 cube.

I'm getting points but not many, i'm pretty bad at this.
[Image: sigimage.php?u=696078&t=212997&b=luna2]
Reply
Likes:
#14
Before you submit your next few cubes, take a screenshot of them and post them here. Perhaps I can point a few things out.

What does your accuracy tell you? Is your inaccuracy more from over or under colouring?
[Image: sigimage.php?FAHUser=hiigaran&FAHTeam=21...hangeling1]
Reply
Likes:
#15
Ugh.. I just can't figure out this eyewire thing...

Sometimes I get 3 points and sometimes I get 80

Meh... the ETA still says 1.5 days... usually it only takes about 5min to figure it out and its been over an hour and a half

I guess I'll wait a few more hours and if it doesn't change then... idk
Reply
Likes:




SOON