基于GoogLeNet的静态图像中人体行为分类研究

Study on Human Behavior Classification in Still Images Based on GoogLeNet

在对GoogLeNet模型分析的基础上,通过Caffe平台上使用开源GoogLeNet模型,对Stanford40静态图像集中人体行为进行分类研究,得到top-5准确率为50.23%,这些工作对深入理解GoogLeNet模型和静态图像中人体行为分类的研究有所帮助。

In this paper, based on the analysis of the model GoogLeNet, we propose to use GoogLeNet model which is open on the Caffe platform to classify the human behavior in the still images of Stanford40, and get the top-5 accuracy 50.23%. These work is helpful for us to deeply understand the model GoogLeNet and the study on human behavior classification in still images.