Foundation Models

Foundation models are large, pre-trained models that can be used on their own, or as part of a vision project, to solve a computer vision problem.

You can deploy the following foundation models on your own hardware with Inference:

• Gaze (LC2S-Net): Detect the direction in which someone is looking.

• CLIP: Classify images and compare the similarity of images and text.

• DocTR: Read characters in images.

• Grounding DINO: Detect objects in images using text prompts.

• Segment Anything (SAM): Segment objects in images.

• CogVLM: A large multimodal model (LMM).

To learn how to deploy these foundation models, refer to the Roboflow Inference documentation.