Vehicle Video Graphic Content Detection With Sensitivity Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current dashcam video analysis systems require manual review of extensive footage for graphic content detection, leading to mental health risks for reviewers and inefficiencies, while automatic systems consume significant computing resources and pose privacy and ethical concerns.
Innovation Solution
A video system that uses object detection, sensitivity scoring, and a multi-modal large language model (MMLLM) to automatically detect and categorize graphic content, reducing human exposure and conserving resources by preprocessing video data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual review of dashcam footage is used for graphic content detection, then detection accuracy is improved, but reviewer mental health is harmed and processing efficiency deteriorates
Solution Approach 1:
The patent introduces an automatic content detection system as an intermediary between the video footage and human reviewers. This system uses machine learning models to pre-screen footage and flag only potentially graphic content for human review, thereby protecting reviewers from direct exposure to harmful content while maintaining detection accuracy through a two-stage process
Solution Approach 2:
The system performs preliminary automatic detection and filtering of graphic content before it reaches human reviewers. By pre-identifying and flagging potentially graphic segments using automated analysis, the system prepares the review process in advance, allowing reviewers to focus only on flagged content rather than reviewing all footage, thus reducing mental health impact while maintaining efficiency
2Measurement precision
If manual review of extensive video footage is used, then comprehensive content analysis is achieved, but processing time and resource consumption increase significantly
Solution Approach 1:
The patent segments the video review process into distinct phases: automatic pre-screening phase and manual review phase. The system divides the extensive footage into segments, automatically analyzes them using machine learning models, and only passes flagged segments to human reviewers. This segmentation dramatically reduces processing time while maintaining comprehensive analysis through the combined strength of both automated and manual phases
Solution Approach 2:
The system applies partial action by having automated models perform the initial screening of all footage, requiring human reviewers to act only on a subset of flagged content. This approach uses excessive automated analysis to cover all bases, then focuses human effort only where needed, reducing overall processing time while maintaining comprehensiveness
3Productivity
If automatic graphic content detection systems are deployed, then processing efficiency is improved, but computing resource consumption increases
Solution Approach 1:
The patent segments the computing workload between automated machine learning models and human reviewers. The system uses lightweight automated detection to process all footage initially, then only processes flagged segments with more resource-intensive manual review. This segmentation improves overall efficiency while controlling computing resource consumption by avoiding full automated analysis of all content
4Loss of information
If comprehensive video footage is made accessible, then information availability is improved, but privacy and ethical concerns worsen due to potential unauthorized dissemination of graphic content
Solution Approach 1:
The patent introduces an automatic content detection system as an intermediary layer between the video storage system and users. This intermediary automatically screens all footage, flags potentially graphic content, and controls access accordingly. This allows comprehensive video footage to remain accessible for legitimate purposes while preventing unauthorized dissemination of graphic content through automated filtering and access control mechanisms
Data Source
AI summary
A device may receive video data associated with a vehicle experiencing an event, and may determine object data identifying bounding boxes, tracks, and labels for objects in the video data. The device may calculate sensitivity scores indicating a likelihood that a person inside the vehicle is injured, a likelihood that a person outside the vehicle is injured, a likelihood that an animal is injured, or a dangerousness of the event, and may aggregate the sensitivity scores to generate an aggregated score. The device may horizontally concatenate a subset of frames of the video data to generate an input image, and may generate queries about whether the video data contains graphic content. The device may process the input image and the queries, with a multi-modal large language model, to determine whether the video data contains graphic content, and may perform actions when the video data contains graphic content.