Bounding Box Predictions
Explore how YOLO predicts bounding boxes by dividing the input image into grid cells and estimating coordinates relative to each cell. Understand coordinate formats, confidence scores, class predictions using logistic regression, and YOLO's multi-scale approach to detect objects of varying sizes.
We'll cover the following...
What is a bounding box?
A bounding box is simply a rectangle drawn around an object to identify the exact location of the object in an image. In OD tasks, it also helps us identify what kind of object is present in an image.
How are coordinates represented?
Mathematically, a bounding box is represented as a tensor consisting of information related to the location of the object and confidence scores. In OD tasks, two formats are widely followed to represent location:
(
, , , ): They are also known as top-left and bottom-right coordinates. (
, , , ...