基于增强型选择性内核网络的幻灯片切换检测

主要针对多种类型的演讲视频进行幻灯片切换检测,其主要研究工作包括:提出了一种基于选择性内核模块和全局上下文模块的增强型选择性内核网络,用于进一步提高演讲视频幻灯片切换检测的准确性。选择性内核模块通过分裂、融合、选择三个部分,调整卷积核权重,提取幻灯片语义信息。全局上下文模块则采用压缩扩充模块和非局部模块,调整通道权重、突出有效信息,又捕获长距离依赖、提取幻灯片的全局信息,增强选择性内核网络看清全局的能力。实验结果表明,增强型选择性内核网络用于检测幻灯片切换,其准确性进一步提升,并较好地解决幻灯片之间发生微小变化时的漏检问题。

This paper mainly studies the detection of slide transition of speech video based on deep learning.The main research work includes:this paper proposes an Improved Selective Kernel Network based on the selective kernel module and the global context module for slide transition detection of lecture videos.The selective kernel module adjusts the weight of the convolution kernel by splitting,merging,and selecting three parts,and extracts the slide semantic information.The global context module uses compression and expansion modules and non-local modules to adjust channel weights,highlight effective information,capture long-distance dependencies,and extract global information of slides,improving the ability of selective kernel networks to see the whole world.Experimental results show that the Improved Selective Kernel Network can better solve the problem of missed detection when small changes occur between slides.