Research On Recommendation Algorithm Based On Photo Scene Recognition

In recent years,user sentiment analysis plays an increasingly important role in personalized recommendation system.A variety of mobile applications in vertical scenes greatly enrich people's ways of social interaction,entertainment and learning.It has become a new social way for contemporary people to share and record their lives by taking photos on their mobile phones and then matching them with a suitable copy.The analysis of users' emotion based on the specific scene of photo and text can provide more accurate personalized recommendation service based on emotional needs.But at present,the analysis of users' emotion in the photo scene is still less.The existing image description generation task is to describe the photo content objectively,which lacks the consideration of emotional factors.Therefore,multimodal user sentiment analysis combined with photos and text is the main content of this paper.Therefore,this paper mainly studies the system design and implementation of"photo and text recommendation learning app".First of all,this paper studies the current situation of user emotion portrait,multimodal emotion analysis and its recommendation field.On this basis,it analyzes the system requirements,completes the functional modules and database design.Then,it constructs user emotion portrait according to the photographing situation information,and adds image metadata features on the basis of the current better image emotion analysis model In order to capture the emotional tendency of users,a scene analysis model is trained to obtain the content of the photos.At the same time,in the case of integrating the recognition of photographic knowledge points and the characteristics of the scene,the system provides users with the service of pushing and learning photographic knowledge points.Finally,this paper uses pychar and Android studio tools as the development tools of the background and client respectively.The deep learning model involved is built and trained through the python framework,and deployed in the back end of flash.The system realizes the functions of recommendation,visualization of user preferences and push of photography knowledge.