Right now, when a model is trained in browser, we have a PREDICT button, which brings you to a menu that lets you select an uncertainty threshold before pushing predictions onto your image. However, in a more useful design, the predictions should not be automatically pushed, but instead highlighted so that a user can manually inspect the predictions, make changes to them, and then push the changes. This can be accomplished with a combination of the CELL tab on the right and the new cellSelectionCanvas or a variant of it for highlighting cells of interest, but some additional UI and/or logic will need to be written for the full workflow.
Right now, when a model is trained in browser, we have a PREDICT button, which brings you to a menu that lets you select an uncertainty threshold before pushing predictions onto your image. However, in a more useful design, the predictions should not be automatically pushed, but instead highlighted so that a user can manually inspect the predictions, make changes to them, and then push the changes. This can be accomplished with a combination of the CELL tab on the right and the new cellSelectionCanvas or a variant of it for highlighting cells of interest, but some additional UI and/or logic will need to be written for the full workflow.