Detect Objects in Images with Ultralytics YOLO

What you’ll get

The ability to point an AI at an image and have it find and label the objects in it (a bus, people, a car), drawing boxes around each. This is “object detection”, a kind of “computer vision” (AI that understands images). YOLO is a fast, popular model for it, and Ultralytics packages it so you can run it in a couple of commands. It is intermediate: you use Python and a terminal, but the model is already trained, so you are using it, not building it.

Tools you need

• Python (free): the language you run YOLO with (version 3.8 or newer).
• PyTorch (free): the AI framework YOLO is built on; it gets installed alongside.
• Ultralytics YOLO (freemium): the library with ready-to-use YOLO models. Free for most uses; commercial use may need a licence.

Steps

1. Install Python and PyTorch: get Python from python.org if you do not have it. Then install PyTorch following the picker at pytorch.org/get-started/locally/ (it gives you the exact command for your system). Check with python --version.

2. Install Ultralytics: this one command pulls in YOLO and what it needs:

   pip install ultralytics

   You should see packages install and finish with a success message.

3. Detect objects from the command line: run a pre-trained model on a sample image straight from your terminal. (yolo26n.pt is a small, fast model; source is the image to analyze.)

   yolo predict model=yolo26n.pt source='https://ultralytics.com/images/bus.jpg'

   It downloads the model if needed, processes the image, and saves a copy with boxes drawn around the detected objects, telling you where it saved it. Open that image to see the results.

4. Detect objects from Python (more control): for use in your own program, save this as detect_objects.py and run python detect_objects.py:

   from ultralytics import YOLO
   
   # Load a pre-trained model
   model = YOLO("yolo26n.pt")
   
   # Run detection on an image (swap the URL for a local file path if you like)
   results = model("https://ultralytics.com/images/bus.jpg")
   
   # Open a window showing the image with the detected objects boxed
   results[0].show()

   A window should pop up showing the image with labelled boxes around each detected object.

Original source

Adapted from the official Ultralytics YOLO GitHub repository, which documents and provides examples for their computer-vision models.

Notes & variations

• Free-tier alternative: no powerful computer? Run the same Python in a free cloud notebook like Google Colab, which does the heavy lifting on Google’s machines (mind the session limits).
• Common pitfall: too little memory or no compatible GPU makes large models slow or causes errors. Start small and make sure your setup meets PyTorch’s basic requirements.
• Tip for better results: start with the small yolo26n.pt for quick tests, then try larger models (like yolo26s.pt or yolo26m.pt) for more accuracy at the cost of speed and memory. For your own images, set source to the file path, for example source='my_image.jpg'.