Saliency Prediction for Adaptive Content Delivery
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional approaches fail to effectively identify and enhance salient points of interest in content items, such as virtual reality content, that are likely to be relevant to users, leading to inefficient content delivery and presentation.
Innovation Solution
A saliency prediction model is trained to predict salient points of interest using aggregated heat map data, which are then enhanced by increasing video quality, zoom level, and blurring non-salient regions, allowing for improved resource allocation and user engagement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional content delivery approaches are used, then all regions of content items are presented uniformly, but resources are wasted on non-salient regions and salient regions do not receive sufficient quality enhancement
Solution Approach 1:
The patent applies local quality enhancement by identifying salient points of interest within content items and selectively enhancing only those regions. The system determines salient points using saliency prediction models and then applies quality enhancement, zoom, or other processing specifically to those identified regions rather than uniformly processing the entire content item. This resolves the contradiction by allocating resources efficiently to only where needed.
2Manufacturing precision
If uniform quality enhancement is applied to all regions of content items, then all regions are presented at high quality, but bandwidth and processing resources are excessively consumed
Solution Approach 1:
The system selectively enhances quality only for salient regions identified through saliency prediction, rather than applying uniform enhancement across the entire content item. This allows high-quality presentation where users are most likely to focus attention while minimizing bandwidth consumption in non-salient regions.
Solution Approach 2:
The patent applies partial action by enhancing only a portion of the content item (the salient regions) rather than the entire content. This selective approach achieves the necessary quality enhancement for user engagement while avoiding excessive bandwidth consumption that would result from uniform enhancement of all regions.
3Ease of operation
If salient points of interest are identified and enhanced, then user engagement is improved, but device complexity increases due to saliency prediction models and processing
Solution Approach 1:
The system performs preliminary saliency prediction and identification of points of interest before the actual content delivery or presentation. By pre-processing the content to identify salient regions, the system can then apply enhancement only where needed during delivery, improving user engagement while managing complexity through staged processing.
Solution Approach 2:
The patent introduces saliency prediction models and processing systems as intermediaries between the content source and the user. These intermediaries analyze and identify salient regions, then guide the enhancement process. While this adds system complexity, it enables significant improvements in user engagement through targeted quality enhancement.
Data Source
AI summary
Systems, methods, and non-transitory computer-readable media can determine saliency information describing one or more salient points of interest that appear during presentation of a content item, wherein the salient points of interest are predicted to be of interest to one or more users accessing the content item and embed the saliency information describing the salient points of interest into the content item, wherein the saliency information is capable of being processed during presentation of the content item to enhance the presentation of the content item.