Challenge Results
These are all the challenges results that have been posted on Mozak.
- Date: May 24, 2017 UTC
- Consensus length: 60068 (81%) Gold Standard length: 74051 This consensus is another great starting point. There was only one section of missed dendrite. The axon was not fully captured, as it could have been with better comments from me and more time posted. There was also some false positive captured at the top, but that was easily identified and removed.
- Date: May 24, 2017 UTC
- Consensus length: 72138 (90%) Gold Standard length: 80155 Other than some missed axon, this consensus is nearly perfect. It is very accurate, and has almost zero false positive. More like this one please!
- Date: May 24, 2017 UTC
- Consensus length: 95542 (106%) Gold Standard length: 90272 This neuron has great overall quality. However, the ambitious Mozak players traced a lot of false positive throughout the image. I think this was one that we left posted for longer than normal, which might account for the false positive. They also captured all the dendrite accurately and almost all of the axon (missing the long stretches at 12 o?clock from the soma). Apart from the false positive, this was a good starting point for us.
- Date: May 24, 2017 UTC
- Consensus length: 92636 (65%) Gold Standard length: 141759 The major challenge for this reconstruction is the neuron?s axon density. The consensus was able to capture all but one of the dendrites, and there is very little false positive, so I think the 65% is fairly accurate for the percentage of total axon that was captured. I do not see evidence of a faster reconstruction by our annotator. I think this is mostly due to the fact that though many branches were captured in the consensus, it is still necessary to check them all and clean up minor errors, which takes about as much time as tracing them from scratch.
