The x, y, width and height pixel coordinates marking where a detected object sits in an image; object detection outputs them, whereas image classification doesn't.
Read more: Microsoft Learn
In the Ultra Transcenders books
Each book explains Bounding box in context, with comparison tables and the common traps.
Terms in this definition
- WHERE
Limits a SELECT, UPDATE or DELETE to just the rows meeting a condition. Omit it, and the statement hits every row.
- Object detection
Vision technique that, for every instance of an object, gives back a label, a confidence score and bounding-box coordinates, making it possible to find and count several objects or classes in an image.
- Image classification
A computer vision technique that gives the whole image one or more labels without locating anything, for example distinguishing benign from malignant or identifying a bear's species.
Related terms
- Area of interest
Smart-crop capability of Image Analysis that, for a requested aspect ratio, returns a bounding box around the most important part of the image, useful for thumbnails and cropping.
- Brand detection
Recognises the logos of thousands of well-known brands from a built-in catalogue, giving back each name with a confidence and bounding box; available in Image Analysis 3.2, and your own logos call for a trained custom model.
- Semantic segmentation
A computer vision method that assigns a class to each pixel, yielding a mask of an object's precise outline rather than just a bounding box.