ROI Video Preprocessing for Blockchain-Verified AI Inferences
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video data streaming practices relay the entire frame of view, which is resource and time-consuming due to high resolution and large data transfer, and there is a need to authenticate video data to resolve disputes about AI model inferences.
Innovation Solution
A system that preprocesses video data to extract regions of interest, applies AI models, and records the data on a blockchain for authentication, ensuring unalterable proof of inferences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If the entirety of streaming video data (entire frame/field of view) is relayed to the consumer, then complete visual information is provided, but data transfer time and resource consumption increase significantly
Solution Approach 1:
The video frame is segmented into multiple regions of interest (ROIs) based on detected objects or areas of significance. Only these segmented regions are extracted and transmitted to the consumer, rather than the entire frame. This segmentation approach maintains essential visual information while dramatically reducing the data volume that needs to be transferred and processed.
Solution Approach 2:
The system extracts only the relevant portions (regions of interest) from the complete video frame. By taking out and isolating the specific areas containing objects of interest, the system eliminates unnecessary data transmission while preserving the critical visual information needed by the consumer.
2Measurement precision
If high resolution video data is transmitted, then image quality is improved, but processing resources and bandwidth consumption increase
Solution Approach 1:
The system applies different quality levels to different regions of the video frame. High-resolution data is transmitted only for regions containing objects of interest, while other areas are either excluded or transmitted at lower resolution. This local quality approach optimizes the balance between image quality and resource consumption by concentrating processing resources where they are most needed.
3Reliability
If AI models process the entire video frame, then comprehensive analysis is achieved, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary processing by detecting and identifying regions of interest before the main AI analysis is applied. By pre-processing the video frame to locate significant areas, the system prepares the data in advance, allowing subsequent AI models to focus their computational efforts only on these identified regions rather than analyzing the entire frame, thus improving processing speed while maintaining analysis reliability.
Data Source
AI summary
A method of committing video data and associated metadata to a blockchain ledger includes preprocessing the video data, by a first computer processor, according to preprocessing parameters defined within a runtime configuration file, wherein the video data is preprocessed to edit the video data to be suitable for use by a first artificial intelligence model. The method also includes accessing the preprocessed video data and associated metadata by the first artificial intelligence model with the first artificial intelligence model being configured to determine a first output that is indicative of a first inference dependent upon the preprocessed video data, formulating first annotated video data and associated metadata by the first artificial intelligence model that includes the first output indicative of the first inference, and accessing the preprocessed video data and associated metadata by a blockchain subscriber independent from the first artificial intelligence model. In response to a determination by the first artificial intelligence model that the first inference has occurred so as to prompt the formulation of the first output, the method includes committing a first entry to the blockchain ledger that includes the first annotated video data, metadata associated with the first annotated video data, the preprocessed video data, and metadata associated with the preprocessed video data.