Posted On: Nov 23, 2022
Amazon Rekognition Labels is a machine learning-based image and video analysis service that can detect objects, people, text, scenes, and activities. Starting today, customers get multiple improvements and enhancements in Amazon Rekognition Labels for images. In the new update, we have added 600 new labels and improved the accuracy of over 2,000 existing labels. We also introduced Image Properties, for image quality and color detection. Lastly, we have added an ability to filter API results by labels and label categories.
Among the many new labels and label categories, customers can now detect, for example, popular landmarks such as Brooklyn Bridge, Colosseum, Eiffel Tower, Machu Picchu, Taj Mahal; activities such as Applause, Cycling, Celebrating, Jumping, Walking Dog; and sports labels such as Baseball Game, Cricket Bat, Figure Skating, Rugby, Water Polo. Customers can use the new Image Properties feature to detect dominant color of the entire image, image foreground, image background, and objects with localized bounding boxes. Image Properties also measures the sharpness, brightness and contrast of the image. Using Image Properties, customers can filter out low quality images or add color metadata to search content. Image Properties is an optional feature priced separately from general labels detection and is only available with the updated AWS SDKs.
Labels API response now contains “aliases” and “categories”. Aliases are other names for the same label and categories group individual labels together based on 40 common themes, such as Food and Beverage and Animals and Pets. Note: Aliases and categories are only returned with the updated AWS SDKs.