MathWorks offers AI-assisted image and video labeling tools in its Computer Vision Toolbox, using the Segment Anything Model (SAM 2) and Grounding DINO to automatically generate pixel-level labels for training deep learning models. These tools are available to licensed users of the toolbox.

In the Image Labeler and Video Labeler apps, users can invoke SAM 2 to automatically segment objects in images or videos by clicking or drawing regions of interest, or by segmenting an entire image automatically. Separately, Grounding DINO allows users to create bounding box labels for object detection by typing descriptive text (e.g., 'car' or 'person') rather than drawing manually. The output is labeled ground truth data used to train or evaluate deep learning models. These capabilities require the Image Processing Toolbox and Deep Learning Toolbox licenses.