|
|
|
AUTHOR(S): Soumya Ashwath, Dr. Jyothi Shetty
|
|
TITLE |
 PDF |
|
ABSTRACT Recent advancements in text-to-image synthesis are explored through innovative approaches designed to address key challenges in generating realistic images from textual descriptions. These approaches include IIR-Net, CRD-CGAN, GALIP, Transformer-based methods, StyleGAN-T, and OPGAN. Each model introduces distinct techniques such as Image Information Removal (IIR), attention mechanisms, CLIP integration, and object-centric architectures, aiming to enhance fidelity, diversity, semantic consistency, and object modelling accuracy. Evaluation across diverse datasets demonstrates their superior performance over existing methods, highlighting improvements in editability, photorealism, and control over the synthesis process. Furthermore, future research directions are discussed, emphasizing the need for refining text alignment, advancing object modelling techniques, and exploring personalized GAN approaches to further advance text-to-image synthesis. |
|
KEYWORDS Image synthesis, Machine learning, Image generation, GAN, Generative models, CGGAN, Object modelling |
|
|
|
Cite this paper Soumya Ashwath, Dr. Jyothi Shetty. (2024) Advancements In Text-To-Image Synthesis: A Comprehensive Review Of Techniques, Challenges, And Future Directions. International Journal of Signal Processing, 9, 16-27 |
|
|
