Building your first computer vision model just got easier

We're thrilled to announce a new repository that makes it incredibly easy for anyone to get started for free with Groundlight, a computer vision (CV) platform powered by natural language. This first steps guide is designed to walk you through setting up your first Groundlight detector to answer a simple question: "Is the door open?" Behind the scenes, Groundlight will automatically train and deploy a computer vision model that can answer this question in real time. With our escalation technology, you don't need to provide any labeled data - you get answers from your first image submission.

What is Groundlight?

Groundlight offers a truly modern take on computer vision, combining the best AI models with real-time human supervision in the cloud. Our Escalation Technology automatically chooses the best solution for your problem - whether that's a traditional CV model like an EfficentNet on the edge, a powerful Visual LLM in the cloud, or a live sensible human monitor. The result is fairly incredible if you're used to the traditional ["gather data", "train model", "evaluate", "repeat"] pattern of machine learning. Instead, Groundlight empowers you to just phrase the visual question you want answered in English, send in images, and Groundlight provides confidence-calibrated answers. At first, the answers will be slow and/or unconfident, but after not very many examples you're using a customized CV model trained just for your task.

What's Inside?

Our getting started repository provides an easy to understand Python codebase that you can run on any modern computer (including a Raspberry Pi). It captures images from a camera of your choice (by default your webcam) and uses Groundlight to continuously train and deploy a computer vision model that determines if your door is open or closed. Whether you're just starting out or a seasoned developer, this example is crafted to provide a smooth introduction to integrating Groundlight into your projects and provide a springboard for building advanced applications with Groundlight.

The main.py file could hardly be simpler. First you just initialize the camera and your Groundlight detector:

camera = setup_camera()

gl = Groundlight()

query_text = "Is the door open? This includes if the door is only partially open."

detector_name = "door_open_detector"

detector = gl.get_or_create_detector(
    name=detector_name,
    query=query_text,
)

and then a simple infinite loop to send images from the camera to your detector:

try:
    while True:
        image = camera.grab()

        image_query = gl.ask_ml(detector=detector, image=image)
        
        print(f"The answer to the query is {image_query.result.label.value}")

        sleep(10)
finally:
    camera.release()

‍

How do I get started?​

Visit the repository and follow the steps in the README. After trying it out, we encourage you to modify the code to solve a real world problem you experience. Doing so should be as simple as changing the query you ask Groundlight. See the Learning More - Additional Resources section of the README for more information. If you want to learn more about the Groundlight Python SDK, which is used to power this repository, check out our SDK or visit the documentation.

We're Here to Help!​

Got questions? We're eager to assist! Reach out to us through email (support@groundlight.ai), or chat on the Groundlight web app - a Groundlight engineer or scientist is available to help every weekday during business hours.

We can't wait to see what you build with Groundlight!