Zero-Shot Object Detection

Type: huggingface.object_detection.ZeroShotObjectDetection

Namespace: huggingface.object_detection

Description

Detects objects in images without the need for training data. image, object detection, bounding boxes, zero-shot, mask

Use cases:
- Quickly detect objects in images without training data
- Identify objects in images without predefined labels
- Automate object detection for large datasets

Properties

Property Type Description Default
model hf.zero_shot_object_detection The model ID to use for object detection {'type': 'hf.zero_shot_object_detection', 'repo_id': 'google/owlv2-base-patch16', 'path': None, 'variant': None, 'allow_patterns': None, 'ignore_patterns': None}
image image The input image for object detection {'type': 'image', 'uri': '', 'asset_id': None, 'data': None}
threshold float Minimum confidence score for detected objects 0.1
top_k int The number of top predictions to return 5
candidate_labels str The candidate labels to detect in the image, separated by commas ``

Outputs

Output Type Description
output List[object_detection_result]  

Metadata

Browse other nodes in the huggingface.object_detection namespace.