What is Faster R-CNN and how does it improve object detection?

Faster R-CNN is a two-stage object detector that uses a Region Proposal Network (RPN) for efficient and accurate detection.

Faster R-CNN improves earlier R-CNN variants by introducing a Region Proposal Network (RPN) that shares convolutional features with the detection network. First, a backbone CNN extracts feature maps. Then the RPN generates candidate object regions (anchors) and scores them. Finally, ROI pooling extracts fixed-size features for classification and bounding box regression. This architecture significantly reduces computation by eliminating external proposal methods like Selective Search.