Saving state while using box prompting

As you know we can train box prompting (ai labeling) while we are labeling images. However when we stopped and close the browser we need to train box prompting again if we want to continue. Can we save the state and continue from last state.

1 Like

Hi @yrg23 - love this feedback!

We’re working on the ability to save these models, should be out soon.

One question for you: if we made the model API-accessible, would you consider using it in production?

Hi Jacob,

I probably would. I’m an academician. This is my first time using Roboflow and I have to say that it really speeds up my labeling process. If I was doing this all the time I would definitely use it, especially with the saving of the model I am fine-tuning, I can label all the data very quickly.

From my little experience with Roboflow, I can say that box prompting is much better for labeling, including self-supervised models. Box prompting also allows me to evaluate in a practical way whether the model I will train later will be capable of generalization. Further development of this feature would definitely motivate me to use roboflow using api or for a fee :slight_smile:

Kind regards…

1 Like

@Jacob_Witt any update on this feature? It would be amazing to have this while labeling. Will speed labeling process across sessions.

Hi @anon123 - we’re currently working on a flow where you can use an Instant model in Auto Label; I believe that will solve a lot of the associated workflows of wanting to re-use a Box Prompting model.

Basically, you would:

  1. Label ~10-50 images with Box Prompting
  2. Add those images to your dataset
  3. This would trigger training of an Instant model
  4. You can then use the Instant model to bulk-label images