Roboflow Changelog: February 2021

Welcome to our monthly roundup post of new features and enhancements. You can find the previous changelog here.

The biggest new features released this month were transfer learning and Label Assist.

Roboflow Organize

Announced teams for all (now teams of up to 3 people can collaborate on the Starter Plan for free).
Promoted the Upload API from beta to general availability (now included free on all plans).
Made it possible to invite collaborators to your teams (and for them to accept or reject those invitations).
Added an "Upload via Email" option (which also means you can now upload training data from anywhere via Zapier).
Launched an object count histogram visualization in the dataset health check.
Implemented support for importing annotations in the Unity Perception synthetic data format.
Added support for importing Tensorflow pbtxt labelmaps.

Roboflow Train

Integrated transfer learning ("Start from Checkpoint") with the ability to start training from a previously trained model's weights.
Added COCO weights as a transfer learning option.
Enable auto-reload of Roboflow Train credits for Pro subscriptions.

Roboflow Infer

Demo'd a preview of our new in-browser edge inference.
Added a "download sample app" option to the "Try with Webcam" demo page.
Launched an image snapshot tool for collecting new training data from the webcam preview page.
Decreased the download size of weights files by 4x.
Enabled browser caching of weights files.

Roboflow Annotate

Released Label Assist (model-assisted labeling using your Roboflow Train models)
Added a COCO-trained model for Label Assist on the Starter plan.
Launched a Label Assist class selector so you can select specific classes to label.
Surfaced confidence thresholding controls for Label Assist.
Released a "Mark as Null" feature.

Other