GAN Discriminator Learning via Vector Space Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing generative adversarial network (GAN) learning methods require significant manual labor for training data preparation and labeling, which is inefficient and labor-intensive.
Innovation Solution
A learning system that includes a discrimination image acquisition module, a discrimination vector calculation module, and a discriminator learning module to automate the learning process by acquiring and calculating vectors for anchor, positive, and negative discrimination images, reducing the need for manual labeling and improving efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling of training data is performed, then the GAN can be trained with sufficient accuracy, but the labor required becomes enormous
Solution Approach 1:
The system uses the GAN itself to generate the training data and labels automatically. The generator creates synthetic images with controlled features, and the discriminator automatically provides labels based on feature changes, eliminating the need for manual annotation while maintaining training quality
Solution Approach 2:
The system pre-processes the training data by generating anchor images and calculating discrimination vectors before actual training begins. This preliminary preparation automates the labeling process and organizes data in advance, reducing the labor burden during the main training phase
2Ease of manufacture
If pre-trained models are used for training data preparation, then the learning process is simplified, but the labor for data preparation remains enormous
Solution Approach 1:
The GAN system performs self-preparation of training data by generating synthetic images and automatically creating labels through feature-based discrimination. This self-service approach eliminates dependency on pre-trained models and manual data preparation while maintaining simplicity
Solution Approach 2:
The system introduces an intermediary labeling mechanism where the discriminator's feature-based output serves as automatic labels. This intermediary process bridges the gap between raw data and training requirements, simplifying the overall process without requiring pre-trained models or manual intervention
Data Source
AI summary
A learning system, comprising at least one processor configured to: acquire, for each feature of a generative adversarial network (GAN) which allows a user to control a plurality of features relating to a generated image, an anchor discrimination image, a positive discrimination image, and a negative discrimination image; calculate, for each feature space corresponding to each of the plurality of features, based on a discriminator of the GAN, an anchor discrimination vector, a positive discrimination vector, and a negative discrimination vector; and execute learning of the discriminator such that, in the feature space corresponding to each of the plurality of features, the anchor discrimination vector and the positive discrimination vector approach each other, and the anchor discrimination vector and the negative discrimination vector become distant from each other.