Controlling the Latent Space of GANs through Reinforcement Learning: A Case Study on Task-based Image-to-Image Translation
Year: 2024
Type: conference-paper
Abstract: Generative Adversarial Networks (GAN) have emerged as a formidable AI tool for the purpose of generating realistic outputs based on training datasets. However, the challenge of exerting control over the generation process of GANs remains a significant hurdle. In this paper, we propose a novel methodology aimed at addressing this issue by integrating a reinforcement learning (RL) agent in conjunction with a latent-space GAN (l-GAN), thereby facilitating the generation of desired outputs. To substantiate the efficacy of our approach, we conducted a series of experiments employing the MNIST dataset, with the inclusion of arithmetic addition as an illustrative task. The outcomes of these experiments serve to validate our methodology. We achieved 95.31% accuracy performing addition task on MNIST dataset.
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4398186455