laughing-q/labelimg

LabelImg is a desktop application that lets you draw boxes around objects in images and label them. If you're building a computer vision system, say, training software to recognize cats, cars, or defective products, you need labeled training data. This tool makes that process straightforward: you load an image, draw rectangles around the things you want to identify, assign labels to them, and the tool saves your work as structured data your AI model can learn from. The app is a straightforward visual editor. You open a folder of images, use your mouse to draw rectangular boxes on each image, type in what each box represents (like "dog" or "broken part"), and hit save. The annotations get stored as text files in standard formats that machine learning frameworks expect. You can work through dozens or hundreds of images this way, and the tool includes keyboard shortcuts to speed things up, for instance, pressing "d" moves to the next image, and "w" starts drawing a new box. People use this when they need to prepare training data for object detection tasks. A manufacturer might use it to label defects in product photos. A self-driving car company would use it to label pedestrians, vehicles, and road signs. A researcher building a wildlife monitoring system would label animals in camera trap footage. It's designed for anyone who needs to manually identify and mark up what's in images before feeding that data into machine learning systems. One note: the original creator is no longer actively developing this project, but it's still available and functional. The maintainers recommend checking out Label Studio, a newer tool from the same community that handles a wider range of data types (text, audio, video, and more) in addition to images. If you only need to annotate images and prefer a lightweight desktop tool, the original LabelImg still works and can be installed via a single pip command.