Computer-Assisted Object Tagging in Digital Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual tagging of digital photographs is time-consuming, especially when dealing with multiple people in a single image, leading to low adoption rates due to the labor-intensive process.
Innovation Solution
A system that automatically detects objects-of-interest in digital images, determines tentative tags based on likelihoods, and presents them to the user for confirmation or correction, streamlining the tagging process by focusing on objects-of-interest only.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual tagging is performed to ensure accurate tag assignment, then tagging precision is improved, but time consumption increases significantly
Solution Approach 1:
The system performs preliminary automated tag assignment using object detection and machine learning models before user review. This preliminary action pre-processes the tagging task, reducing the time users need to spend while maintaining accuracy through subsequent user verification of the pre-assigned tags.
Solution Approach 2:
The patent replaces manual mechanical tagging operations with automated computer vision and machine learning systems. Object detection algorithms and ML models automatically identify and assign tags to objects in images, substituting the manual mechanical process with automated intelligent systems that reduce time consumption while maintaining or improving precision.
2Productivity
If automated tag assignment is implemented to reduce time consumption, then productivity is improved, but tag assignment accuracy deteriorates
Solution Approach 1:
The system implements a feedback mechanism where automatically assigned tags are presented to users for review, correction, or confirmation. User feedback on automated tag assignments is used to refine and improve the accuracy of future automated tagging, creating a continuous improvement loop that maintains high productivity while ensuring tag assignment accuracy.
Solution Approach 2:
The patent introduces an intermediary review step between automated tag generation and final tag assignment. Users act as intermediaries who verify and correct automated tags, bridging the gap between fast automated processing and accurate final results, thus maintaining both productivity and accuracy.
3Measurement precision
If all objects in an image are presented for tagging, then completeness of tagging is improved, but user workload increases
Solution Approach 1:
The system extracts and presents only the most relevant or uncertain tag assignments to users for review, rather than requiring users to review all possible tags for all objects. This extraction of critical items reduces user workload while maintaining tagging completeness through selective focus on areas needing human verification.
Solution Approach 2:
The patent applies partial action by having users review only a subset of tag assignments (those with lower confidence scores or higher importance) rather than all tags. The automated system handles the majority of tagging tasks, while user intervention is applied partially to critical cases, reducing overall workload while maintaining completeness.
Data Source
AI summary
One embodiment of the present invention provides a system that facilitates computer-assisted tagging of objects in a digital image. During operation, the system receives locations for one or more objects-of-interest in the digital image. Next, the system determines likelihoods of specific tags being assigned to the objects-of-interest. The system then automatically assigns tentative tags to the objects-of-interest based on the determined likelihoods. Next, the system displays the assignments of tentative tags to a user, and receives corrections to the assignments, if any, from the user.