It is our pleasure to announce that CoralNet is out of Beta and on to “CoralNet 1.0”.
It has now been nearly 10 years since we launched the site and you might ask “why now?”. Well, over the years we have worked to make it more robust and user friendly, and we finally felt ready to drop the Beta. In particular there are a few recent key changes.
- 1) We released a new computer vision backend this month. Confer this blog post for all the details, but in summary the new backend reduces classification error rate by 22.4% while running several times faster! Old sources will stay on the old backend unless you explicitly toggle this on the admin page while new sources will default to the new backend (see screenshot for details). Both work with the API. We will continue to support the old backend, but we recommend everyone to migrate to the new one at their convenience.
-
2) We recently released an API that anyone can use to deploy a classifier they have developed on the site to large amounts of survey images. This is a critical step in automating the full workflow from image capturing through annotation towards analytics. This blog post has the details.
-
3) We have completed several rounds of software and system updates to make the site more robust and maintainable. A lot of this work was completed as part of the recent grant from NOAA. For all the details confer the GitHub repository project page. A big part of the clean-up is a refactoring of our computer vision backend. The code for that is also public.
-
4) We addressed two significant annotation tool display bugs, one related to zooming in, and one related to image orientation. This blog post has the details.
As always, please bring ideas and suggestions. We also welcome contributions through joint grants, in-kind, or other venues.
Happy coralnetting!