Simple model sharing
Models are now able to be shared across all md.ai domains to test, validate, and more! While defaulting to private, the model card allows users to select options to allow export or cloning of models if desired. For example, a user can share a model to validate on another user’s dataset without revealing their model code, and without having to obtain the other users dataset.
Clone model output as annotation
Speed up labeling with Artificial Intelligence! You can now run one of MD.ai’s helper AI models, or your own, and easily copy the model output to use as annotations.
Easy model export
Export your model through the side card or model card for running locally if desired. Detailed instructions are available here.
Better project management with QA role
Added new quality assurance (QA) user role to simplify project management. A user assigned to a QA role can read, write and change their annotations as well as the annotations of other users. The ability to change other users’ annotations as needed represents additional privileges from a regular write user but does not allow the full project access available to administrators. A list of all user types is available here.
Simplify education projects with metadata labels
New flexibility allows for both global and local metadata labels as well as the ability to toggle metadata label visibility. For example, in a synchronous classroom learning scenario the teacher can easily hide answer key metadata label annotations by toggling off the metadata label, and then reveal the answers to all students with one click after the test period is complete.
Label during cine
Speed up labeling of cine images by applying series or exam scoped global labels while cine is playing.
Additional log in options
MD.ai supports log in with GitHub, Google, LinkedIn, or Microsoft as well as institution SSO so you have one less password to remember. You can use that extra brain space for your next great project!
New models deployed
Try out or share some of the many new models available this month including the first place model from the SIIM-FISABIO-RSNA COVID-19 Detection challenge deployed here.
Resolved unwanted zoom during cine which occurred during viewport size change, such as the collapse of a side panel.
Changed email provider for MD.ai account verifications so the email is less likely to be filtered into spam folders.
Resolved bug that caused sign on error if one user was using multiple environments.
Resolved issue where rearranging labels rapidly would lead to delay in visual rearrangement.
Fixed cine keyboard shortcut.