Text Prompts for Image Annotation with Grounding DINO

Overview

Grounding DINO is a zero-shot object detection model that combines DINO transformer architecture with grounded pre-training. This fusion results in a model that can do bounding box image classification tasks based on corresponding text prompts.

Benefits

• Zero-shot Detection: Detect almost anything with natural language.
• High Performance: Rapid object detection COCO zero-shot 52.5 AP (training without COCO data!). COCO fine-tune 63.0 AP.
• Configurable: Flexible ML backend allows you to configure models based on your performance requirements, including the ability to swap object detection models (COCO) out for image segmentation models (SAM).

Using the power of Grounding DINO with Label Studio gives users a powerful tool for improving annotation efficiency using natural language prompts. You can build your own Label Studio integration or contribute to the further development of ML integrations by reading more here.