Grounding DINO is a super cool model for text-based object detection. This means that, given an image and some text queries, the model is able to automatically detect the queries in the image without requiring manual labeled data.

It is similar to OWL-ViT and OWLv2.

2 notebooks are included here, one that showcases inference with Grounding DINO and another one which combines Grounding DINO with SAM in order to generate masks based on text prompts.