YOLO-World Model

Run a zero-shot object detection model.

Run YOLO-World, a zero-shot object detection model, on an image.

YOLO-World accepts one or more text classes you want to identify in an image. The model returns the location of objects that meet the specified class, if YOLO-World is able to identify objects of that class.

We recommend experimenting with YOLO-World to evaluate the model on your use case before using this block in production. For example on how to effectively prompt YOLO-World, refer to the Roboflow YOLO-World prompting guide.

Type identifier

Use the following identifier in step "type" field: roboflow_core/yolo_world_model@v1 to add the block as a step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
class_names	List[str]	One or more classes that you want YOLO-World to detect. The model accepts any string as an input, though does best with short descriptions of common objects..	✅
version	str	Variant of YoloWorld model.	✅
confidence	float	Confidence threshold for detections.	✅

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds YOLO-World Model in version v1 has.

Input and output bindings

• input

  • images (image): The image to infer on..

  • class_names (list_of_values): One or more classes that you want YOLO-World to detect. The model accepts any string as an input, though does best with short descriptions of common objects..

  • version (string): Variant of YoloWorld model.

  • confidence (float_zero_to_one): Confidence threshold for detections.

• output

  • predictions (object_detection_prediction): Prediction with detected bounding boxes in form of sv.Detections(...) object.

Example JSON definition

{
	    "name": "<your_step_name_here>",
	    "type": "roboflow_core/yolo_world_model@v1",
	    "images": "$inputs.image",
	    "class_names": [
	        "person",
	        "car",
	        "license plate"
	    ],
	    "version": "v2-s",
	    "confidence": 0.005
	}