Use Grounding DINO, Segment Anything, and GPT-4V to label images with segmentation masks for use in training smaller, fine-tuned models.
Here is an example of SAM-GPT-4V used to label the manufacturer of a car:
![Screenshot 2023-11-07 at 15 05 44](https://private-user-images.githubusercontent.com/37276661/283588679-6e166715-216e-4712-bd2f-d6c48526b305.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.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.xtymBdKg1F828CzeFVdtZS-IGmv6IXqQqtClD-nXA4c)
This project is licensed under an MIT license.