摘要
网页内容监测是维护网络安全的重要技术手段,针对网站中存在的大量违规图片,如色情、暴力、垃圾广告图片,文章提出一种基于深度学习的网页违规图片检测方法。通过自建多类违规图片数据集,并提出一个基于MobileNet的轻量级网络模型,同时实现色情、暴力和广告三类违规图片的检测。该违规图片检测模型支持一次检测多类违规图片,检测速度快,准确率较高,可以为中小型网站提供违规图片的实时监测服务。
Web content monitoring is an important technical means to maintain network security.Aiming at a large number of illegal pictures in the website,a web page illegal picture detection framework based on lightweight deep convolution neural network is proposed.Different from the single class violation image detection service provided by a single interface in the industry,this paper constructs a multi class violation image data set,and proposes a network model based on MobileNet to realize the detection of multi class violation images at the same time.The illegal picture detection model supports the detection of multiple types of violation images at one time,with fast detection speed and high accuracy,and can provide illegal picture detection services for small and medium-sized websites.
作者
余聪
李柏岩
刘晓强
Yu Cong;Li Baiyan;Liu Xiaoqiang(College of Computer Science and Technology,Donghua University,Shanghai 201620)
出处
《现代计算机》
2022年第13期45-50,共6页
Modern Computer
