Image Recognition Device Using Segmented Category Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image recognition techniques often result in incorrect object recognition due to the use of general object identifiers that recognize all objects, leading to inefficiencies and reduced accuracy, as they may incorrectly identify objects not actually present in the video.
Innovation Solution
An image recognition device that sets specific sections within a video, uses a general object identifier to recognize objects in one section, selects a sub-category for objects in another section, and applies an individual object identifier to improve recognition accuracy by limiting the scope of recognition and reducing false positives.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a general object identifier is used to recognize all objects in a video, then the coverage of object recognition is improved, but the recognition accuracy deteriorates due to incorrect identifications
Solution Approach 1:
The patent segments the broad category of objects into multiple sub-categories (e.g., dividing 'vehicles' into 'cars', 'trucks', 'buses'). The selection unit selects appropriate sub-categories based on video content, and individual object identifiers are applied within these narrower categories, improving accuracy while maintaining comprehensive coverage.
Solution Approach 2:
Different identification strategies are applied to different parts of the video content. The selection unit analyzes video characteristics and selects appropriate sub-categories and individual object identifiers for specific sections, allowing locally optimized recognition accuracy while maintaining overall system versatility.
2Measurement precision
If multiple individual object identifiers are used for different sub-categories, then the recognition accuracy is improved, but the device complexity increases
Solution Approach 1:
The system dynamically selects which individual object identifiers to use based on the video content and detected objects. The selection unit activates only the necessary identifiers for the current video section, reducing computational complexity while maintaining high recognition accuracy when needed.
Solution Approach 2:
The selection unit serves multiple functions: it categorizes objects, selects appropriate sub-categories, and chooses individual object identifiers based on video content. This multi-functional approach consolidates what would otherwise require separate systems, reducing overall device complexity.
3Measurement precision
If manual labeling is performed to add object names to videos, then the recognition accuracy is improved, but the time consumption increases significantly
Solution Approach 1:
The system performs automatic object recognition and labeling without human intervention. The setting unit, first recognition unit, selection unit, and second recognition unit work together to automatically identify objects, select appropriate categories, and generate labels, eliminating the need for manual labeling while maintaining high accuracy.
Solution Approach 2:
The system performs preliminary object identification and category selection before final labeling. The first recognition unit pre-identifies objects and the selection unit pre-selects appropriate sub-categories, which streamlines the final labeling process and enables efficient automatic operation.
Data Source
AI summary
An image identifying device includes: a setting unit which sets a section having at least one image in a video; a first recognizing unit which calculates a plurality of feature amounts related to at least the one image and which acquires a plurality of identification results corresponding to each of the feature amounts from an identifier which may identify a plurality of objects belonging to a first category; a selecting unit which selects, based on the identification results, a second category of a third category; and a second recognizing unit which calculates another feature amount related to an image included in another section and acquires another identification result corresponding to the feature amount from another identifier which may identify the objects included in the second category.