Image File Metadata for Subject Detection and AI Inference Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems fail to effectively manage the information detected from input images, such as subject detection results, which are crucial for applications like automation and autonomous driving.
Innovation Solution
An image processing apparatus and method that includes a detection unit, inference unit, and generation unit to detect subjects, perform inference processing, and generate image files by combining image data, subject information, and inference model information, with the inference model recorded in a private area of the image file.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If subject detection and AI inference results are processed separately from image data, then processing flexibility is maintained, but information management efficiency deteriorates
Solution Approach 1:
The patent merges image data, subject detection information, and AI inference results into a single integrated image file structure. The detection information is stored as metadata within the image file itself, creating a unified data structure that improves information management efficiency while maintaining processing flexibility through standardized metadata formats.
Solution Approach 2:
The patent introduces a standardized metadata structure as an intermediary between the image data and AI inference results. This metadata layer acts as a mediator that organizes and connects different data types, enabling efficient information management without creating complex direct relationships between all data components.
2Reliability
If inference models are stored externally from image files, then model updates are simplified, but association between image data and inference results deteriorates
Solution Approach 1:
The patent embeds the inference model information and results directly within the image file structure, creating a nested organization where the image data contains both the original image information and the AI processing results. This nesting ensures reliable association between image data and inference results while keeping the overall structure manageable through hierarchical organization.
3Ease of operation
If detection information is not recorded in the image file, then image file simplicity is maintained, but usability in subsequent workflows deteriorates
Solution Approach 1:
The patent segments the image file into distinct sections: the original image data and a metadata section for storing detection information and inference results. This segmentation allows the image file to maintain its core functionality while adding new capabilities, improving workflow usability without excessively increasing data volume through structured organization.
Data Source
AI summary
An image processing apparatus comprises: a detection unit that detects a subject from an image; an inference unit that performs inference processing using an inference model on the detected subject; and a generation unit that generates an image file by combining image data of the image, information of the subject, an inference result of the inference processing, and information of the inference model. The generation unit records the inference model in a private area of the image file.