Anchor boxes vs predicted bounded box

I am not able to understand the difference between the anchor boxes and the predicted bounded box. Are they the same or not? can someone explain their uses separately??

The number of anchor boxes and their shapes is determined during exploratory analysis on the data set. They are inputs to running the algorithm. They are static and known in advance of running an image through the pipeline, whether training or prediction time. The number of anchor boxes influences the output shape of the network. The shape of the anchor boxes influences the efficiency of the localization training (but not of the classification training).

Bounding boxes are derived from the predicted object center location and shape outputs of forward propagation, either at training or prediction time. They are dynamic, and depend on the specific image being processed, as well as the thresholds applied in output post processing (eg non-max suppression).

There are a few extended explanations already in the forum which you should be able to find using search. Here is one …