Neural Network Image Coloring with Confidence-Based Querying
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for producing webtoon content are inefficient, particularly in coloring, which hinders the ability to keep up with the rapid expansion of the online content market and increasing user demand, leading to delays and inaccuracies in content delivery.
Innovation Solution
A method utilizing an artificial neural network to generate intermediate images, color queries, and user inputs to progressively complete the coloring of images, allowing for iterative refinement and automatic coloring, while also querying users for unclear regions to ensure accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual coloring is performed by producers, then coloring accuracy is maintained, but production time increases significantly
Solution Approach 1:
The system enables self-service automatic coloring through neural networks that automatically color images without requiring manual producer intervention for routine coloring tasks, while still allowing producer oversight for quality control
Solution Approach 2:
A query system acts as an intermediary between automatic coloring and manual coloring by identifying uncertain regions and selectively querying producers only for those specific areas, rather than requiring complete manual coloring or accepting potentially inaccurate automatic coloring
2Productivity
If automatic coloring is performed using neural networks, then production speed increases, but coloring accuracy decreases in uncertain regions
Solution Approach 1:
The system applies automatic coloring partially - only to regions where the neural network is confident about the coloring decision. Regions with low confidence are identified and set aside for manual review, avoiding excessive automatic coloring that would reduce accuracy
3Measurement precision
If producers prepare all scenes individually, then coloring quality is maintained, but content supply speed cannot keep up with market expansion
Solution Approach 1:
The coloring process is segmented into automatic coloring for confident regions and manual coloring for uncertain regions identified by the query system, allowing parallel processing and optimizing both speed and quality
Data Source
AI summary
A method of completing coloring of an image based on a query regarding a color-unknown region in the image and an answer to the query, includes: generating, by using an artificial neural network, a first intermediate image in which at least one uncolored region in a primary image is colored; generating, by using the artificial neural network, a first color query regarding the at least one color-unknown region in the primary image; and generating a secondary image based on at least one of the first color query, a first answer to the first color query, and the first intermediate image.