Right now sequential frames each need to have the object's bounding box drawn on. Adding in a lightweight model that quickly trains on only the previously labelled images would be a super nice feature that would improve the usefulness of this tool even more.
Right now sequential frames each need to have the object's bounding box drawn on. Adding in a lightweight model that quickly trains on only the previously labelled images would be a super nice feature that would improve the usefulness of this tool even more.