-
This was a test to see if the quantized groundingdino model see:
saburq/groundingdeno_model_quant_int8works. -
Now moved to offical version:
onnx-community/grounding-dino-tiny-ONNX
Install grounding-dino version of transformers.js
git clone git@github.com:decision-labs/test-groundingdino-quantized-js.git
cd test-groundingdino-quantized-js
git clone git@github.com:huggingface/transformers.js.git
cd transformers.js
git checkout add-grounding-dino
npm run build
cd ..Install dependencies.
Make sure you are in the root of the repo.
yarn installRun server
npx nodemon app.jsRun client
python draw_boxes.py
open output_with_boxes.pngIn draw_boxes.py change the text to another label like "tree".
http://localhost:3000?model_name=object-detection&text=tree&image_uri=https://content.satimagingcorp.com/static/galleryimages/Satellite-Image-Paris-Pont-des-Arts-bridge.jpg
