Image Style Conversion Transfer Training with Limited Samples
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image style conversion models face reduced accuracy due to the high cost and difficulty in obtaining large quantities of high-quality training samples, particularly for tasks like animation generation.
Innovation Solution
An image style conversion method that enhances low-quality samples through quality enhancement and feature extraction, utilizing a full style conversion model for migration training to create a target style conversion model, enabling style conversion with limited low-quality samples.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large quantities of high-quality training samples are obtained through manual creation, then the accuracy of image style conversion model training is improved, but the cost and time consumption increase significantly
Solution Approach 1:
The method performs preliminary quality enhancement on target style images before they are used for training. By pre-processing the images to improve their quality, the system reduces the need for extensive manual creation of high-quality training samples, thereby decreasing data preparation time while maintaining training accuracy.
Solution Approach 2:
The patent introduces an image quality enhancement model as an intermediary between low-quality target style images and the style conversion model training. This intermediary process improves the quality of training images automatically, eliminating the need for time-consuming manual image creation while ensuring sufficient training quality.
2Manufacturing precision
If manual creation of uniform-style training pictures is performed, then the quality of training data is improved, but the cost and complexity of data preparation increase
Solution Approach 1:
The system enables self-service data preparation by automatically enhancing the quality of target style images through a trained quality enhancement model. This eliminates the need for manual intervention in creating uniform-style training pictures, reducing both cost and complexity while maintaining high training data quality.
Solution Approach 2:
The patent replaces the mechanical process of manual image creation with an automated image quality enhancement process. By using computational methods to enhance image quality instead of manual drawing or editing, the system significantly reduces preparation complexity and cost while achieving the same or better training data quality.
3Productivity
If limited low-quality samples are used for training, then the cost and time of data preparation are reduced, but the accuracy of model training deteriorates
Solution Approach 1:
The method changes the quality parameter of training images through automated quality enhancement. By applying the quality enhancement model to improve low-quality samples, the system transforms limited low-quality inputs into effective training data, maintaining model training accuracy while improving data preparation efficiency.
Data Source
AI summary
Embodiments of this application disclose an image style conversion method performed by an electronic device. The method includes: performing quality enhancement on a first target style image to obtain a second target style image; performing feature extraction on the second target style image to obtain a target style feature; performing migration training on a preset target style conversion model by using a full style conversion model and the target style feature to obtain a target style conversion model; inputting a full style feature, the target style feature, and a to-be-converted image into the target style conversion model, and performing style conversion on the to-be-converted image using the target style conversion model to obtain a target image conforming to a target style.