awesome-openai-vision-api-experiments

Why we included this project

Roboflow maintains this collection of working experiments built on the OpenAI Vision API, each with runnable code and, in several cases, a hosted demo you can open in a notebook or Gradio app. The examples lean practical: simple image classification, zero-shot detection that pairs GPT-4V with GroundingDINO or Segment Anything, comparisons against CLIP, and a webcam chat demo. Alongside the code, the repo links the papers and blog posts behind each technique, so you can see the reasoning before you build. That makes it a handy starting point for developers weighing whether a vision-language model.