Hierarchical Video Brand Safety Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in automatically determining the brand safety of videos to assess whether they are suitable for associating with commercial content, given the vast and complex nature of video data, which often includes inappropriate content such as nudity, violence, or tobacco, requiring a method to analyze audio, metadata, and image frames at various levels of granularity.
Innovation Solution
A video brand safety determination technique is implemented using machine learning models to extract frame-level, scene-level, and video-level features, combining deep learning and natural language processing to generate a brand safety score, indicating whether it is safe to associate a brand with the video by analyzing visual, audio, and textual content at different hierarchical levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single model is used for brand safety analysis, then the device complexity is low, but the measurement precision and accuracy of brand safety analysis is insufficient
Solution Approach 1:
The video content analysis system is segmented into three distinct hierarchical levels: frame-level models analyzing individual video frames, scene-level models analyzing groups of frames构成 scenes, and video-level models analyzing the complete video. Each level uses specialized models trained for specific tasks, enabling comprehensive brand safety analysis while maintaining manageable complexity through modular design
Solution Approach 2:
The system transitions from analyzing videos at a single level to a multi-dimensional hierarchical approach. By adding temporal dimension (frame-level) and contextual dimension (scene-level and video-level), the system achieves more accurate brand safety determination through analysis at multiple scales simultaneously
2Measurement precision
If comprehensive multi-level analysis is performed on video content, then the brand safety analysis accuracy is improved, but the processing time and computational resources increase
Solution Approach 1:
The system performs preliminary analysis at the frame-level first, extracting basic features and identifying potential brand safety issues in individual frames. This preliminary action prepares data in advance for the subsequent scene-level and video-level analysis, enabling efficient multi-level processing without redundant computations
Solution Approach 2:
By dividing the video into frames and scenes, the system can process content incrementally at different levels of detail. Not all videos require full multi-level analysis - the segmented structure allows selective application of analysis depth based on content characteristics, reducing overall processing time
Data Source
AI summary
Disclosed herein are techniques for determining brand safety of a video including image frames and audio content. In some embodiments, frame-level features, scene-level features, and video-level features are extracted by a set of frame-level models, a set of scene-level models, and a set of video-level models, respectively. Outputs from lower level models are used as inputs for higher level models. A brand safety score indicating whether it is safe to associate a brand with the video is determined based on the outputs from the set of video-level models. In some embodiments, commercial content associated with the brand is insert into the video that is determined to be safe for the brand.