Image Text Generation Accuracy via User Correction Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to accurately generate and provide relevant text for groups of images, leading to inadequate memory and storytelling of events captured in photographs.
Innovation Solution
An image processing apparatus and method that includes a text generating device, correction accepting device, and updating mechanism, which uses a storage device to improve text accuracy by referencing subject combinations and user corrections across multiple images, and updates the text generation method based on user feedback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic text generation is performed for image groups, then text production efficiency is improved, but text accuracy deteriorates
Solution Approach 1:
The system implements a feedback mechanism where users can correct generated text, and these corrections are stored and used to retrain the text generation model. This allows the system to maintain high productivity while continuously improving accuracy through user feedback loops.
Solution Approach 2:
The system enables self-service by automatically generating text for image groups without requiring manual input, while still allowing users to correct errors. The system uses stored correction data to automatically improve its generation quality over time, reducing the need for continuous manual intervention.
2Measurement precision
If text generation method is updated using user corrections, then text accuracy is improved, but device complexity increases
Solution Approach 1:
The system performs preliminary action by storing user corrections and preparing training data in advance. This allows the text generation method to be updated systematically without adding immediate complexity to the operational workflow, as the data preparation is done beforehand.
Solution Approach 2:
The system creates a copy of the text generation model that incorporates user corrections. Instead of modifying the original model directly, the system generates updated versions based on stored correction data, which simplifies the update process and reduces complexity by allowing model versions to be independently managed.
Data Source
AI summary
Provided are an image processing apparatus, an image processing method, and an image processing program capable of relatively easily creating text representing a description of an image. From a plurality of images selected from an image group, text describing the plurality of images is generated according to a text generation method. Generated text is displayed in a first text display region and a second text display region. In a case where text is corrected by a user, the text generation method is updated. More accurate text is generated.