Image Comment Generation Using Depth-Based Subject Positioning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image management systems struggle to create accurate comments for images that include subjects other than 'person', as they fail to effectively utilize the positional relationships between multiple subjects to describe the situation depicted in the image.
Innovation Solution
An image management apparatus and method that acquires data and positional information from images, computes the relative positional relationships between subjects in the depth direction, and uses this information to create comments that accurately describe the image scenario, including combinations of subjects like people and buildings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If the subject attribute is restricted to 'person' for comment generation, then the system complexity is reduced and processing is simplified, but the accuracy and completeness of image description deteriorates when multiple subject types (e.g., people and buildings) are present
Solution Approach 1:
The system changes the parameter of subject attribute classification from a single category ('person') to multiple categories (people, buildings, objects, etc.). This allows the comment generation system to accurately describe images containing various subject types by identifying and processing different subject attributes separately, thereby improving image description accuracy without significantly increasing system complexity.
2Productivity
If only positional information of people is used for comment creation, then the processing speed and efficiency are improved, but the comprehensiveness of situation description deteriorates when buildings or other objects are involved
Solution Approach 1:
The system applies a universal comment generation process that handles multiple subject types (people, buildings, objects) through the same computational framework. The computation unit computes relative positional relationships for any combination of subjects, and the creating unit generates comments based on these relationships, making the system multi-functional and comprehensive without sacrificing processing efficiency.
Solution Approach 2:
The system adds the dimension of subject type classification to the existing positional information processing. By considering not only the position of subjects but also their attributes (person, building, object), the system creates more comprehensive situation descriptions while maintaining efficient processing through structured attribute-based computation.
3Measurement precision
If detailed positional relationships between multiple subjects are computed, then the accuracy of situation description is improved, but the computational load and processing time increase
Solution Approach 1:
The system segments the comment generation process into distinct stages: acquiring positional information, computing relative positional relationships, and creating comments based on subject attributes. This segmentation allows for optimized processing at each stage, computing positional relationships only when needed and using pre-defined comment templates to reduce overall processing time while maintaining high accuracy.
Data Source
AI summary
Provided is an image management apparatus including: an acquiring unit configured to acquire data of an image and positional information on a subject included in the image; a computation unit configured to compute a relative positional relationship between a plurality of subjects included in the image in at least a depth direction, based on the acquired positional information on the plurality of subjects included in the image; and a creating unit configured to create a comment data relevant to the image automatically, based on the computed relative positional relationship between the plurality of subjects.