基于Caffe网络模型的Faster R-CNN算法推理过程的解析

Analysis of Faster R-CNN Algorithm's Inference Based on Caffe Modeling Framework

Faster R-CNN是基于深度学习的目标检测算法,在PASCAL VOC的目标检测竞赛中表现优异,是当前一个研究热点。从该算法对应的网络模型出发,明晰网络模型中每一层输入输出的数据维度和主要参数,在此基础上,解析算法的推理过程,把握算法的技术关键,从而为后续的深入研究提供一种新的思路。

Recently, Faster R-CNN is a research hotspot of object detection algorithms based on deep learning, which has excellent performance in the competition of the PASCAL VOC Challenge （PASCAL VOC stands for Pattern Analysis, Statistical modeling and Computational Learning Visual Object Classes）. Aimed at the modeling framework of Faster R-CNN, attempts to illustrate data dimensions and major parameters of each layer＇s input and output. By this means, makes an insight to the inference and key points of Faster R-CNN. and brings out a novel thinking for further research.