New to roboflow’s UI (have been loving your python libraries) and I had a question about Auto Label settings.
When I run a test, it gives me the option to adjust the confidence threshold, which defaults to 50%.
It used to have an option to “get code” (still in the docs here) to run autodistill locally, but I can’t see this anymore.
My main question is: as Grounding DINO has two confidence thresholds (box and text), what does the 50% refer to? 50% for both or just one? Basically I want to run the code locally using the default 50% threshold and would like to know what parameters I should use it running locally.
The confidence threshold you see in the Roboflow UI refers to the level of certainty the model has in its predictions. By default, the model will return predictions when it is 50% confident source. This threshold can be adjusted to reduce false-positive predictions.
As for Grounding DINO, it’s not explicitly mentioned in the extracts which of the two confidence thresholds (box and text) the 50% refers to. However, Grounding DINO is an object detection model, so it’s likely that the confidence threshold applies to the bounding box predictions it makes.
Regarding running Autodistill locally, the extracts do mention that Roboflow provides all the necessary code to auto-label images on your own device and upload the resulting annotations back to Roboflow source. However, the specific code for this is not provided in the extracts.
For more detailed information on how to use Autodistill locally, you might find the Autodistill Documentation useful. It provides a method label() that you can use to label your images. The method includes parameters like record_confidence which might be relevant to your use case.
I hope this helps! Let me know if you have any other questions.
Best,
kapa.ai
Hi! I’m Roboflow’s AI powered bot. This is an automated response powered using AI with knowledge from our docs, blog, and forum. If this solves your issue, please mark it as the solution using the checkbox icon. If this doesn’t solve your issue, please follow up with what you’ve tried/why, and the community will continue to chime in as usual.
