from Hacker News

Show HN: Simple demo tool for Gemini 2.0 Flash object detection

by PetrBrzyBrzek on 12/17/24, 1:06 PM with 4 comments

I built a straightforward demo to showcase Gemini's object detection capabilities. Upload any image, get bounding boxes and clean JSON output. It's currently the only model I've found that can accurately return object bounds.

Demo: https://langtail.com/gemini-bounding-boxes

The tool is pretty simple:

- Drop an image or use example ones (apples, llamas)

- Hit detect

- Get visual bounding boxes + JSON output

Happy to answer any questions!

  • by Jefro118 on 12/18/24, 12:05 PM

    Just been playing around with the bounding box feature myself - do you mind sharing how you figured out how to translate the bounding boxes back to the correct dimensions/position on the original image?

  • by Alifatisk on 12/17/24, 1:16 PM

    Why Gemini 2.0 Flash? Any specific reason?