Multimedia Content Object Recognition for In-Video Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video processing technologies fail to enrich video-related functions to meet diverse user requirements, leading to a limited presentation area that affects user experience.
Innovation Solution
A method and apparatus that recognize target resource objects in multimedia content and determine corresponding recommended objects based on predetermined correlations, presenting them to users in a card set with scrolling functionality, enhancing interaction and consumption paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional video processing technology is used, then the system structure remains simple, but the video-related functions are not enriched and user experience is limited
Solution Approach 1:
The patent segments video content into multiple dimension tags (subject matter, time, location, etc.) and processes each dimension independently through dedicated recognition modules. This segmentation allows the system to enrich video functions without creating a monolithic complex structure, as each tag type can be processed by specialized components working in parallel.
Solution Approach 2:
The patent creates a universal video processing framework that handles multiple types of resource objects (items, music, locations, people) through a common architecture. The same basic flow (recognition→tagging→recommendation) applies to different content types, enabling the system to provide diverse video-related functions while maintaining structural efficiency through reuse of core components.
2Adaptability or versatility
If more video-related functions are added to enrich user experience, then the presentation area becomes limited and affects user experience
Solution Approach 1:
The patent transitions from two-dimensional video playback to multi-dimensional interaction by adding temporal, spatial, and categorical dimensions through tags. Resource objects are recommended based on multiple dimensions simultaneously (e.g., items appearing at specific times and locations), effectively expanding the functional space without increasing the physical presentation area.
Solution Approach 2:
The patent implements nested information structures where video content contains resource objects, which contain tags, which contain multiple attribute dimensions. This nested organization allows rich multi-dimensional functionality to be embedded within the existing video presentation framework, adding depth rather than requiring additional surface area.
3Loss of information
If multiple target resource objects are recognized in multimedia content, then the information completeness improves, but the processing complexity increases
Solution Approach 1:
The patent divides the recognition task into separate modules for different resource object types (item recognition, music recognition, location recognition, person recognition). Each module processes its specific object type independently, reducing the complexity of handling multiple object types simultaneously while ensuring comprehensive information extraction from the video content.
Solution Approach 2:
The patent performs preliminary classification of resource objects by type before detailed processing. By pre-identifying what type of object is being processed (item, music, location, or person), the system can apply type-specific recognition algorithms and tag structures, simplifying the overall processing complexity while maintaining complete information extraction across all object types.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure provides a method, an apparatus, a device, and a storage medium for processing multimedia content. The method includes: in response to a predetermined trigger operation acting on a presentation page of first multimedia content, recognizing at least one target resource object carried in the first multimedia content, where there is predetermined correspondence between the target resource object and a recommended object type; then, determining, based on the predetermined correspondence, a recommended object type corresponding to a first target resource object in the at least one target resource object, and determining at least one recommended object based on the first target resource object; then, presenting the at least one recommended object. According to the embodiments of the present disclosure, in presenting a multimedia content, a recommended object related to the target resource object can be presented to a user based on a target resource object carried in the multimedia content, and an extended consumption path for content carried in the multimedia content is provided to the user, thereby improving the user experience.