Mobile Image Effect Processing With Server-Trained GAN Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Mobile terminals face challenges with slow response rates and poor effect outcomes in image or video processing due to computing power limitations, affecting user experience.
Innovation Solution
Deploy a first generation model on the mobile terminal, trained by a second generation model on a server, using a generative adversarial network, with training data generated by the server to enhance computing efficiency and image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If effect processing is performed on mobile terminal using local computing resources, then processing speed and response rate improve, but processing quality and effect outcome deteriorate due to computing power limitations
Solution Approach 1:
The system segments the effect processing task into two parts: a lightweight first generation model deployed on the mobile terminal for rapid processing, and a comprehensive second generation model on the server for quality enhancement. This segmentation allows the mobile terminal to handle time-critical operations locally while the server handles computationally intensive quality improvement tasks.
Solution Approach 2:
The first generation model acts as an intermediary between the user's effect processing request and the server's second generation model. It provides immediate preliminary results on the mobile terminal while enabling subsequent refinement through server-based processing, thus bridging the gap between speed and quality requirements.
2Manufacturing precision
If comprehensive training data is used to train the first generation model, then model accuracy and effect quality improve, but training time and computational resources increase
Solution Approach 1:
The second generation model is trained comprehensively on the server in advance using extensive training data. This preliminary action creates a high-quality source model that encapsulates learned patterns and effects, which can then be distilled into the lighter first generation model for deployment on mobile terminals without requiring the terminal to perform the time-consuming comprehensive training.
Solution Approach 2:
The first generation model is essentially a compressed or distilled copy of the second generation model's knowledge. Instead of training the first generation model from scratch with extensive data, the system copies the essential processing capabilities and learned patterns from the server-trained second generation model, significantly reducing training time and resource requirements on the mobile terminal.
Data Source
AI summary
The present disclosure relates to an effect processing method, an electronic device and a non-transitory storage medium. The effect processing method includes: in response to an effect processing request input on a mobile terminal, acquiring an image to be processed; and performing effect processing on the image to be processed by a first generation model deployed on the mobile terminal to obtain a target effect image, and displaying the target effect image; the first generation model is obtained by training a first generative adversarial network, at least part of training data of the first generative adversarial network is generated by a second generation model deployed on a server, and the second generation model is obtained by training a second generative adversarial network.