Ground Truth Images and Video
Computer Vision Toolbox™ provides a complete workflow for generating ground truth data from images and videos to train AI models for tasks such as object detection, semantic segmentation, instance segmentation, text recognition, and image or video classification. You can start by using the Image Labeler and Video Labeler apps to interactively annotate data with a wide range of label types. These include rectangles, polygons, polylines, scene labels, and pixel-level labels. To get started labeling a collection of images, see Get Started with the Image Labeler. To get started labeling a video or sequence of images, see Get Started with the Video Labeler.
The Image Labeler and Video Labeler apps support manual, AI-assisted and automated annotation, allowing you to accelerate labeling using built-in AI models like the Segment Anything Model (SAM) and Grounding DINO. For more information, see Get Started with AI-Assisted and Automated Labeling. You can also integrate custom automation algorithms to tailor the labeling process to your specific needs. For more details, see Create Custom Automation Algorithm for Labeling.
Once labeling is complete, you can export the annotated data and postprocess it to create training data sets for AI models. The toolbox supports workflows for organizing and managing labeled data, enabling seamless integration with training pipelines for classification, detection, and segmentation tasks.
For collaborative projects, the Image Labeler app includes features to manage team-based labeling, enabling you to distribute labeling tasks, review annotations, provide feedback, and track progress across multiple contributors. This makes it easier to scale labeling efforts and maintain consistency across large data sets. For more details, see Get Started with Team-Based Labeling.
Highlighted Topics
- Choose an App to Label Ground Truth Data
- Get Started with the Image Labeler
- Get Started with the Video Labeler
- Get Started with AI-Assisted and Automated Labeling
- Get Started with Team-Based Labeling
- Training Data for Object Detection and Semantic Segmentation
- Postprocess Exported Labels for Instance Segmentation Training
Categories
- Label Images and Video
Label images and video using Image Labeler and Video Labeler apps
- AI-Assisted and Automated Labeling
Automate labeling using AI-assisted tools like SAM and Grounding DINO, create custom automation algorithms
- Manage Team Labeling Projects
Create and manage collaborative labeling projects, distribute labeling and review tasks among team members using Image Labeler app
- Use Ground Truth for Training AI Models
Preprocess, augment, and split ground truth data for training and evaluating AI models
Featured Examples
Automatically Label Ground Truth Using Segment Anything Model
Produce pixel labels for semantic segmentation using the Segment Anything Model (SAM) in the Image Labeler app. The SAM is an automatic segmentation technique that you can use to segment object regions to label with just a few clicks, or automatically segment the entire image and instantaneously create labels for selected regions. In this example, you interactively label pixels for semantic segmentation in two ways.
Automatically Label Ground Truth Using Vision-Language Model
Automatically label ground truth images for object detection using the Grounding DINO vision-language model (VLM).
Automate Ground Truth Polygon Labeling Using Grounded SAM Model
Combine Grounding DINO and the Segment Anything Model 2 (SAM 2) to automatically produce polygon labels using the Video Labeler app.
Automate Ground Truth Labeling for Semantic Segmentation
Use a pretrained semantic segmentation algorithm to segment the sky and a road in an image.
Automate Ground Truth Labeling for Instance Segmentation
Create an automation algorithm to automatically label data for instance segmentation using a pretrained SOLOv2 network in the Video Labeler app.
Automate Ground Truth Labeling for Object Detection
Create an automation algorithm to automatically label data for object detection using a pretrained object detector.
Automate Ground Truth Labeling for OCR
Automate the labeling of text for OCR training and evaluation.
Automate Labeling of Objects in Video Using RAFT Optical Flow
Use a pretrained RAFT optical flow estimation network to propagate a predefined object mask from one frame to the next in a video sequence.
Export Ground Truth Object to Custom and COCO JSON Files
Export a ground truth object to a custom data format JavaScript Object Notation (JSON) file and to a COCO data format JSON file.
Convert Image Labeler Polygons to Labeled Blocked Image for Semantic Segmentation
Convert polygon labels stored in a groundTruth object into a labeled blocked image for semantic segmentation workflows.
