Oscar Beijbom
Oscar Beijbom
This needs to be fleshed out, and it is not a huge priority. But it has been requested several times.
Suggestion: how about we add a banner at the top encouraging anyone still using the old features to switch? It could be one of these yellow bubble things that we...
Check out this image: https://coralnet.ucsd.edu/image/646341/annotation/tool/ it is confirmed, but not fully annotated. Ideally, the "confirmed" boolean should be replaced by a property that asserts if all points are annotated. But...
In the annotation tool: be able to re-assign points. Click a point and "re-assign" it. This means that the point is removed and a new one is generated using the...
Right now we ask users to specify the annotation area right from the start. I think it would be nicer to delay this if possible for two reasons 1) We...
This needs more details, but the suggestion is that when a user creates a label, search the previous labels for near-duplicate names, and show them to the user asking if...
Currently, the validation results are stored as blob-data on S3 and loaded as needed when the user moves to see the classifier stats. This isn't robust since it creates dependencies...
I was thinking about image filesize limits. I don't think there is any reason to have limits anymore with the new backend. I'll conduct some tests, and if they work...
From I.W.: "Have you thought much about extending CoralNet capacity to help users diagnose their sources and training quality (eg allow download of user-annotation, robot-annotation, confidence-level for each point?; other...
https://groups.google.com/d/msg/coralnet-users/iQnW_Ftln6Y/plp_b2OCAgAJ From user: "Until now, we have uploaded many photos and defined the number of points to be 25 per photo. Now, we want to switch to 50 or a...