Research On Object Detection And Content Recommendation In Short Video Based On Deep Learning

With visible improvement of the infrastructure and network environment in the era of mobile Internet,consumer's lifestyle and entertainment habits are undergoing tremendous changes.The main contents of the Internet are transitioning from text to pictures or videos.The trend of video mobility is unstoppable,which leads to the vigorous development of short video industry.Short videos are much easier to create,spread and share,with diversified information.This satisfies the user's social needs and meets the user's fragmented time management habits.Short video thus becomes a new fashion."With the increasing user stickiness and industry scale,the short video industry are facing great opportunities of content monetization.Advertising is the most direct form of content monetization.In recent years,advertisement delivery has shown a trend of migration from the PC side to the mobile side.The traditional advertisements are usually stiffly inserted into the videos,which are inefficient and always decrease users' experience.At this stage,it remains a great challenge for short video industry to find new advertising strategies and achieve high rate of return.With the development of computer vision technology,object-level video detection becomes a possibility.This thesis proposes a systematic scheme for video object detection and content recommendation based on the deep learning model Faster R-CNN.This scheme will match the video contents to the displayed advertisements based on the principles of high correlation,precision and low interruption,thus obtains a balance between recommendation and user experience.Two system modes are available according to the video sources and network environments,named as Cloud Mode and Mobile Terminal Mode.The Cloud Mode is composed of a Server,Content Delivery Network(CDN)and Clients.The Server will detect and recognize the contents of the CDN videos in advance,match them to corresponding advertisements by some recommendation algorithms and play the contents on the mobile Clients.The Mobile Terminal Mode mainly processes non-CDN resources such as some local videos,completes the tasks of object detection,recognition and content recommendation with limited computation ability.In both modes,the system can achieve personalized content recommendation based on user behavior statistics.The main research works and contributions of this thesis are as follows:(1)The Cloud Mode successfully implement the functions such as video frame acquisition,key frame extraction,video object detection and recognition with Faster R-CNN and contextual advertising on the Clients.(2)The Mobile Terminal Mode adopts a self-developed player kernel to achieve real time video frame extraction,as well as an advanced Faster R-CNN model with MobileNet to reduce the computational burden.Besides,the system will automatically select the users' most preferred products based on feedbacks and retrain the model with fewer outputs.These improvements ensure the feasibility and efficiency when the system are implemented on mobile terminals.(3)This thesis establishes a complete platform for video object detection and content recommendation,and tests the speed,accuracy and advertising effects of the system in both modes.