Faster R-CNN
A two-stage object detection framework that uses a Region Proposal Network for efficient detection.
Overview
Faster R-CNN, introduced by Ren et al. in 2015, is a landmark object detection architecture that introduced the Region Proposal Network (RPN) — a fully convolutional network that simultaneously predicts object bounds and objectness scores at each position. This replaced the slow selective search used in earlier R-CNN variants.
Key Details
The architecture has two stages: the RPN generates region proposals, then a second network classifies each proposal and refines its bounding box. Faster R-CNN was the first end-to-end trainable object detector and achieved near real-time performance. It remains the foundation for many detection systems, including Mask R-CNN.