It's quite interesting, but some of the "correct" results don't make sense from either a description or sample image stand point.

For example
Baffled
edit: More Baffled

Could someone explain how that is the right answer? asymmetrical bloby/tubular compared to the "correct" symmetrical vainy/web.


Is that "correct" in the sense that an above average number of people on sisi clicked on it, so we're being penalised because people are just clicking through carelessly? a bug? naff data in the project? etc.


edit: couple of samples later, I get something almost the same as before "ok, it was this last time..." wrong.... erm

As one of the researchers for this project I would say that your first image looks like a plasma membrane staining, together with nucleoli (and maybe also an additional cytoplasmic and nuclear staining, hard to tell when not being able to toggle the different channels). The point is however that 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. The descriptions of the different locations are only true if there would be a cytoplasmic staining only. This is a bit tricky of course, but usually the different stainings don't have the exact same intensity which might make it possible to differentiate them.

I agree that the second image would rather be labeled as vesicles, but I can't really tell what the consensus for that location is from the image.

And what we have seen from the result from the test server is that you guys actually are doing really good!

I'm glad it helped!

We will of course keep track of how well you are doing and also keep taking your inputs into account! This is however not an easy task we have been giving you, so I'm expecting the consensus to improve (even though it's way far from bad already, most classifications are fine) with time. It takes a while to learn what to look for and to get a feeling for all the different stainings. Trust me, I'm basing this on my own experience

By request of the HPA guys over at twitter here is a recording of me playing Project Discovery. In the video I explain why I make the choices I make. I hope it gives you guys a frame of reference when doing Project Discovery.

How to classify samples

Well... Need to say i'm feeling the game 'guess community decision' here instead of 'analyze the image and set categories'....
After about 20 images i'm starting to lose interest because i'm bad at social predictions and most of the time community chosen categories are not closer to image than the ones i have chosen.

Thanks for noticing! We are aware of the problem and we have notified CCP about it so hopefully it will be fixed soon.

I am unsure if this has been reported already but I am experiencing a bug with project discovery on this toon. Ever since I started project discovery on this toon (I have another toon that I started it on) I have not been able to see my submission history at all. Also, once I hit 80% or so, the only thing that would affect my accuracy rating was the guaranteed right/wrong samples. I am currently rank 42 with ~95% accuracy.