This is an implementation of Grounding DINO by IDEA Reseach as a Cog model. Grounding DINO can detect arbitrary objects with human text inputs such as category names or referring expressions. The model architecture combines Transformer-based detector DINO with grounded pre-training to achieve open-vocabulary / text-guided object detection. See the paper, original repository and this Replicate model.

Follow the model pushing guide to push your own fork of Grounding DINO to Replicate.

You will need to have Cog and Docker installed on your local to run predictions. You can use Grounding DINO to query images with text descriptions of any object. To use it, simply upload an image and enter comma separated text descriptions of objects you want to query the image for. Expected input arguments are:

• image: your input image
• query: text queries describing objects you want to detect, separate queries with commas
• box_threshold: chooses the boxes whose highest similarities are higher than a box_threshold
• text_threshold: extracts the words whose similarities are higher than the text_threshold as predicted labels

To run a prediction:

cog predict -i [email protected] -i query="a pink mug" -i box_threshold=0.2 -i text_threshold=0.25 

To build the cog image and launch the API on your local:

cog run -p 5000 python -m cog.server.http

@article{liu2023grounding,
  title={Grounding dino: Marrying dino with grounded pre-training for open-set object detection},
  author={Liu, Shilong and Zeng, Zhaoyang and Ren, Tianhe and Li, Feng and Zhang, Hao and Yang, Jie and Li, Chunyuan and Yang, Jianwei and Su, Hang and Zhu, Jun and others},
  journal={arXiv preprint arXiv:2303.05499},
  year={2023}
}