Recommendation and Visual Search

With the advent of Internet and its adoption by billions of users, the volume of online content has skyrocketed to billions of pages, images, videos and more. Leading digital platforms and small marketers as well need to organize their offerings and select those contents which are most likely to interest their users. How do Amazon, ebay, YouTube and others do it? RSIP Vision tells you how recommendation systems work. You too can recommend the best to your users: contact us and we will be happy to discuss!

Video Recommendation

Here is a simple explanation of how a Recommender System work. Take YouTube as an example: a huge quantity of videos needs to be processed, classified, tagged and ranked by users or by an automatic algorithm, before it can be used by the Recommender system. Learn how users are classified by filtering (either collaborative, content-based or hybrid) and videos are ranked by similarity. Read more...

Automatic Video Categorization

Online video hosting platforms utilize a variety of methods for content discovery. Recommender systems allow users to face the huge amount of information offered to their view: recommendation algorithms analyze the video and automatically suggest a confined set of adequate tags to the uploader. Alternatively, the system learns to automatically assign tags to videos without any user intervention. Read more...

Automatic Semantic Tagging

Sites containing huge amounts of content must necessarily recommend only a narrow and relevant list to users. Recommender systems can be seen as tool to automatically generate personalized search preferences, with the purpose of keeping the balance between monetized targeted suggestions and satisfactory user experience. Automatic semantic tagging helps them do just this. Read more...

Automatic Human Action Segmentation

When a video uploader disregards adding tags and categories, online video hosting platforms encounter what is called the new item problem. This can be solved by utilizing visual analysis of videos and images: first, by filtering videos into recognizable objects and combining human action segmentation and recognition; later, by training Multiclass Support Vector Machines to assign labels to detected actions in the temporal domain of videos. Read more...