If an API (Application Programming Interface) could be exposed (possibly as a web service) to allow developers, especially those with skills in AI algorithms, to develop software to perform the mapping function, this could drastically speed up the time required to complete the project.
We have run a competition to enable AI researchers to develop algorithms for automated analysis of serial EM images:
http://fiji.sc/wiki/index.php/Segmentation_of_neuronal_structures_in_EM_stacks_challenge_-_ISBI_2012We are planning to run improved versions of this competition in the future, ideally with monetary prizes.
Perfect! Exactly what I was looking for.
Thanks,
Dean Horak
I was looking for stuff like this a while back to play with, but the datasets were nothing like Eyewire and usually vast (i.e. undownloadable over broadband).
Do you mind if we register just to get the data to play with, or do you have a small stack of similar data on a webserver somewhere?
Edit - there are some referenced at the link below, but only one seems to be ssTEM.
@backupelk the competition is officially over but the data is still available. So please download it. In a sense the competition is still ongoing, as Ignacio will continue to update the leaderboard as people send in their results.
The Open Connectome project http://openconnecto.me has datasets, but I am not sure whether any are annotated (i.e. colored).