Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Analyzing and Improving Latent Diffusion Models
Latent diffusion models (LDMs) [1] have recently emerged as a powerful tool for high-quality image generation, offering superior scalability and training efficiency compared to pixel-space diffusion models. While the network architectures of LDMs have received significant attention, other design aspects of these models (for example the forward noise schedule and the autoencoder) remain underexplored. This project aims to enhance the characteristics of LDMs, e.g., quality and efficiency, by investigating various design elements of latent diffusion models.
Keywords: diffusion models
This project is designed to push the capabilities of latent diffusion models and advance their quality. Since LDMs are the main technology behind many recent breakthroughs in diffusion-based generative models, improving LDMs will have significant practical impact. The exact direction of the project will be decided according to the interests and prior knowledge of the student. Please feel free to reach out for more details.
**References**:
[1] Rombach, Robin, et al. "High-resolution image synthesis with latent diffusion models." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022.
[2] Karras, Tero, et al. "Elucidating the design space of diffusion-based generative models." Advances in Neural Information Processing Systems 35 (2022): 26565-26577.
This project is designed to push the capabilities of latent diffusion models and advance their quality. Since LDMs are the main technology behind many recent breakthroughs in diffusion-based generative models, improving LDMs will have significant practical impact. The exact direction of the project will be decided according to the interests and prior knowledge of the student. Please feel free to reach out for more details.
**References**:
[1] Rombach, Robin, et al. "High-resolution image synthesis with latent diffusion models." Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022.
[2] Karras, Tero, et al. "Elucidating the design space of diffusion-based generative models." Advances in Neural Information Processing Systems 35 (2022): 26565-26577.
The outcome of the project should be publishable in a top-tier conference (e.g., CVPR, NeurIPS, etc.)
The outcome of the project should be publishable in a top-tier conference (e.g., CVPR, NeurIPS, etc.)
