IDEA-Research/Grounded-Segment-Anything ? reverse-engineered prompt

Build me a simple demo app that lets me type what I want to find in an image, then detects those objects and draws clean masks around them. I want to upload a photo, enter a text prompt like “person, cup, dog”, and get the segmented results back right away with bounding boxes and masks overlaid on the image.

If it’s not too much, also add an easy automatic labeling mode so I can use the same tool to create annotations for a dataset and save the results. A small web interface would be great, and if you need to check current setup details online or reuse the existing notebooks and demos, feel free to do that. Keep it practical and easy to run locally, with a straightforward flow for testing on a few images.