I haven’t found a QC thread for this round of Project Discovery or a reporting tool in-game so here goes:
1 Like
Is this where I say that the big “FAILED” notification blocking the most important part of the screen is extremely unhelpful? Because it’s really annoying when I want to see what I missed, but I can’t.
Speaking of unhelpful elements blocking the screen, what’s with those 5x5 blocks in the corners? As far as I can tell, all those do is hide cells for no reason.
7 Likes
(To help thread searchers find this: Project Discovery Citizen Science Covid-19 Flow Cytometry improvement tests)
I agree with the comment re test “fail/succeed” message obscuring the view of the “gold standard” vs. my own selection. Given that this is the feedback we need to improve, it’s pretty important to have an unobstructed view. Seems like an easy fix.
I’ve reached level 30 and have never yet received a test for anything but a single variant of cells. It’s the one with a multicolored one on top and a coarse-grained sparse monochromatic one beneath. (A few times I’ve seen it rotated 90 degrees.) There are several other variants that keep recurring again, and again, and again - but there are no tests for them. It would be so very helpful to have tests, or just samples with the gold standard, showing how to divvy those up. I have no idea if my guesses for those variations are on the mark, or wildly off. Thanks!
3 Likes
1 Like
I agree. Whoever is deciding on the ‘gold standard’ for some of these samples isn’t as gold as they think they are.
4 Likes
After 8 submissions, I am not able to continue. The trick is to get below the submission standard in order to continue it appears.
I can usually gain around 5 levels before it tells me to go away for 22 hrs. It’s really annoying when you’re on a roll.
3 Likes
I think they are trying to slow down the number of Marshall Blue prints that will be flooding the market.
2 Likes
Mayhap so.
I sometimes get a fail or an accuracy %, but what does it mean when it just says “Submitted”?
Theyre done by a computer program in bulk
The ones with gold standards check your accuracy. The others are ones that the reseachers want to be demarcated by us players.
Well, the program is ■■■■.
1 Like
Yes it is, that’s why we’re doing flow cytometry instead of letting a program do all of it
How are we supposed to be helping the fight for covid, when we cannot describe a cluster?
These are not even edge cases. Proj discovery completely ignores centres, as we have been taught.
If the premise is that computer do not understand a cluster, then who is grading the answers?
Shouldn’t their be some learning going on, or consensus.
Let’s kill covid
1 Like
As above, some of the Gold Standard accuracy tests are very obviously wrong.
I also think the method of demarcation is flawed, seems obvious to me that most of the clusters are oval or comet shapes, and many are overlapping. As you cannot overlap the polygons, some of the more complex images are not possibly to delineate correctly. A better tool would be something that allows overlapping ovals?
1 Like
As you are in fact doing research whilst ‘playing’ there are not tests for every image, the tests exist I imagine to check that you have a basic grasp of what you supposed to be doing. I have seen several, but only where there are two clusters, never for more than this.
1 Like
The “gold standard” that splits a cluster in half is clearly INCORRECT and I have not seen that one.
My problem is the training is all on fairly simple cases and then one quickly gets thrown MUCH more complicated cases without ANY feedback and…
Come on… a little feedback here! 8? 9? 2?
Am I just supposed to take a wild ass guess???
2 Likes