Roboflow Changelog: January 2021
Welcome to our monthly changelog where we catalog our recent feature additions and improvements. If you missed it, you can find last month's changelog here.
This month we launched Roboflow Annotate, an integrated image annotation tool included in all Roboflow plans. In the first few weeks, over 1500 users have already cumulatively spent over 2500 hours labeling images!
Roboflow Organize
- Improved speed and reliability for datasets over 100,000 images.
- Added feature to move individual images between train/valid/test set.
- Released annotation histogram on Roboflow Pro's advanced dataset health check.
- Added Scaled-YOLOv4 annotation format.
- Improved video processing (including support for LabelBox video annotations) on Roboflow Pro.
- Published upload API code snippets in several languages.
- Released an enterprise security page.
- Launched new in-app upgrade dialog to clarify plans and pricing.
Roboflow Train
- Included in-app visualization of models' predictions on their datasets' validation and test sets.
- Added (optional) hyperparameter evolution beta for Pro users (reach out if you're interested in trying it out).
Roboflow Infer
- Detailed the philosophy behind our hosted model API.
- Improved inference speed by ~300ms.
- Added ability for enterprise customers to deploy their hosted model APIs to other AWS regions.
- Onboarded several users to our on-device edge deployment private beta (reach out if interested).
- Fixed confidence and overlap thresholds in the sample web app.
Roboflow Annotate
- Launched Roboflow Annotate, our new integrated image annotation tool! 🎉
- Updated our Pricing Page (Roboflow Annotate is included in all plans).
- Added numeric keyboard shortcuts to the class selector.
- Released documentation of the tools and keyboard shortcuts.
- Added a labeling tutorial to our onboarding flow.
Other
- Added a Scaled-YOLOv4 notebook to our model library and wrote a how to train Scaled-YOLOv4 tutorial.
- Collaboration with Von Deon on a limited edition T-Shirt for projects we feature on our blog.
- Added a Press Page.
- 33 bugs fixed.
- 2 new public datasets.
- 12 blog posts.
- 8 YouTube videos.
- Mentions in Nature, BuiltIn, Vice, Augmented Startups, AI Weekly, AppleInsider, Clay and Milk, Towards Data Science, Medium, Jworks.
- Fun user projects: computer vision powered pizza slicer, hand gesture UI.
- Took some time off to celebrate Christmas and New Year's. 🎄🎆