Image Tagging via Object Orientation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image tagging technologies primarily focus on identifying individuals within images and do not effectively provide tags that accurately classify the image itself based on its content, rather than the individuals in it, leading to inefficiencies in image search and classification.
Innovation Solution
An image information processing device that extracts objects from images, calculates their orientation, and provides tags based on this orientation, allowing for accurate classification of image content such as portrait, landscape, or scene types by analyzing the orientation and presence of objects within the image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional face-based tagging technology is used, then individual identification is achieved, but accurate image content classification is not realized
Solution Approach 1:
The patent segments the tagging process into two distinct functions: face detection for individual identification and orientation analysis for image content classification. By separating these functions, the system can simultaneously achieve individual tagging (through face detection) and image content classification (through orientation analysis of detected objects), thereby resolving the contradiction between individual identification and accurate image classification.
Solution Approach 2:
The patent introduces a new dimension of analysis by examining object orientation angles in addition to object detection. Instead of relying solely on face detection results, the system analyzes the orientation of detected objects (faces, bodies, etc.) to determine image content categories such as portrait, landscape, or group scenes. This dimensional expansion enables accurate image content classification while preserving individual identification capabilities.
2Adaptability or versatility
If only face detection is performed, then individual tags are provided, but image-level classification tags are missing
Solution Approach 1:
The patent makes the object detection and orientation analysis mechanisms universal by applying them to multiple purposes: individual identification through face detection, image content classification through orientation analysis, and scene type determination. This multi-functionality ensures comprehensive tagging coverage that includes both individual-level and image-level information, eliminating the loss of classification data while maintaining individual tagging capabilities.
Solution Approach 2:
The patent performs preliminary object detection and orientation calculation on all detected objects before generating tags. By预先 (in advance) analyzing the orientation of detected faces and bodies, the system can determine image content categories and generate appropriate classification tags before final tag provision, ensuring no classification information is lost while maintaining comprehensive tagging coverage.
3Reliability
If conventional tagging methods are used, then person identification is achieved, but image search efficiency is reduced
Solution Approach 1:
The patent implements feedback mechanisms where orientation analysis results continuously refine and supplement face-based identification tags. By analyzing the orientation of detected objects and using this information to generate additional classification tags, the system creates a feedback loop that enhances search efficiency without compromising identification accuracy. The orientation-based tags provide additional search dimensions that improve productivity while maintaining reliable person identification.
Data Source
AI summary
An image information processing apparatus comprising: an extraction unit that extracts an object from a photographed image; a calculation unit that calculates an orientation of the object as exhibited in the image; and a provision unit that provides a tag to the image according to the orientation of the object.