Text Conditioned Image Generation
Text-conditioned image generation aims to create images from textual descriptions, focusing on improving image quality, text alignment, and safety. Current research explores various model architectures, including diffusion models (often enhanced with techniques like consistency models for speed improvements) and autoregressive models, with a strong emphasis on mitigating biases and harmful content generation through methods such as prompt manipulation and latent space correction. This field is significant due to its potential for creative applications and its challenges in addressing ethical concerns related to bias and safety, driving advancements in both generative modeling and responsible AI development.
Papers
November 27, 2024
October 3, 2024
June 10, 2024
May 28, 2024
January 1, 2024
December 8, 2023
November 27, 2023
September 20, 2023
July 10, 2023
June 15, 2023
June 1, 2023
May 30, 2023
May 28, 2023
May 25, 2023
April 28, 2023
March 16, 2023
November 9, 2022
October 28, 2022