Mastering Computer Vision with TensorFlow 2.x

A comparison of YOLO and Faster R-CNN

The similarities between YOLO and Faster R-CNN are shown in the following table:

YOLO	R-CNN
Predicts the bounding box per grid cell.	Selective search generates the bounding box for each region proposal (which is essentially a grid cell).
Uses bounding box regression.	Uses bounding box regression.

The differences between YOLO and Faster R-CNN are shown in the following table:

YOLO	R-CNN
Classification and bounding box regression occur simultaneously.	Selective search generates a bounding box for each region proposal—these are separate events.
98 bounding boxes per image.	About 2,000 region proposal bounding boxes per image.
2 anchors per grid cell.	9 anchors ...