Artificial Intelligence

Faster R-CNN

Faster Region-based Convolutional Neural Network

A deep learning architecture designed for object detection that improves upon the R-CNN by introducing region proposal networks (RPNs) to share computation between the stages of generating region proposals and performing classification and bounding box regression. / 面向目标检测的深度学习架构，通过引入区域建议网络（RPN）在生成区域建议和执行分类与边界框回归阶段之间共享计算，改进了R-CNN。

Knowledge type: 数学理论
First proposed or discovered: 2015
Proposer: Shaoqing Ren, Kaiming He, Ross Girshick, Jian Sun
Creator: 农场主
Last updated: 2026-10-08 07:56:44

Sub-knowledge points

Region Proposal Network (RPN)ROI PoolingMulti-task loss function

Higher-order knowledge points

Deep Learning / 工艺技术 Object Detection / 实际应用Computer Vision

Main parameters and performance indicators

anchor boxes: predefined boxes used to predict object scales and aspect ratiosregion proposal network (RPN): a fully convolutional network that predicts object boundaries and objectness scores simultaneouslyFast R-CNN: the predecessor architecture that introduced ROI pooling

Open problems

Computational complexity during trainingDifficulty in detecting very small objectsHigh memory consumption

Related literature

ranked by community stars

Innovative deep learning computational approach for wheat disease detection employing faster region-based convolutional neural networkOpen source

Faster R-CNN

Sub-knowledge points

Higher-order knowledge points

Main parameters and performance indicators

Open problems

Related products and links