Author(s): Ashik Kumar

Abstract: The generative adversarial network (GAN) framework has emerged as a powerful tool for various image, video, and audio synthesis tasks, allowing the synthesis of visual content in an unconditional or input-conditional manner. It has enabled the generation of high-resolution photorealistic images and videos, which is a more challenging or impossible task than prior methods. It has also led to the creation of many new applications in content creation. In this article, an overview of GANs is provided with a special focus on algorithms and applications for visual synthesis. Also, several important techniques are covered to stabilize GAN training, which has a reputation for being notoriously difficult. Here, GAN’s applications, such as image translation, image processing, video synthesis, and neural rendering are also discussed. Therefore, the proposed model will be developed for Realistic Synthetic Media Generation for Multi-Model, including Images, Videos, and Music using 3GPAN and AFMT-ALKM. Initially, the Audio, Video, and Image datasets will be taken, and pre-processing steps, such as frame conversion, noise removal, contrast enhancement, and background subtraction will be done for three datasets. Next, the object detection will be done by using YOLO, and Motion will be estimated using LET-BMA. Next, the Audio Features, Video Features, Image Features, Appearance Features, Action Features, Emotion Features, and 3D model will be given to the Realistic Media Generation phase for training using 3GPAN. Finally, AFMT-ALKM-based rendering will be done for the retrieved 3D model, and the proposed model will be compared with various models using the result parameters like MSE, RMSE, MAPE, etc.

DOI: 10.33545/27076571.2025.v6.i2b.190

Pages: 139-143 | Views: 93 | Downloads: 58

Download Full Article: Click Here