High-Resolution Image Synthesis with Latent Diffusion Models

This is the seminar presentation of "High-Resolution Image Synthesis with Latent Diffusion Models". A work by Rombach et al from Ludwig Maximilian University of Munich & IWR, Heidelberg University and Runway ML. This paper was accepted at CVPR 2022.
Paper Abstract:
By decomposing the image formation process into a sequential application of denoising autoencoders, diffusion models (DMs) achieve state-of-the-art synthesis results on image data and beyond. Additionally, their formulation allows for a guiding mechanism to control the image generation process without retraining. However, since these models typically operate directly in pixel space, optimization of powerful DMs often consumes hundreds of GPU days and inference is expensive due to sequential evaluations. To enable DM training on limited computational resources while retaining their quality and flexibility, we apply them in the latent space of powerful pretrained autoencoders. In contrast to previous work, training diffusion models on such a representation allows for the first time to reach a near-optimal point between complexity reduction and detail preservation, greatly boosting visual fidelity. By introducing cross-attention layers into the model architecture, we turn diffusion models into powerful and flexible generators for general conditioning inputs such as text or bounding boxes and high-resolution synthesis becomes possible in a convolutional manner. Our latent diffusion models (LDMs) achieve new state-of-the-art scores for image inpainting and class-conditional image synthesis and highly competitive performance on various tasks, including text-to-image synthesis, unconditional image generation and super-resolution, while significantly reducing computational requirements compared to pixel-based DMs.
The source code of the project is publicly available at:
github.com/CompVis/latent-dif...
The full text of the paper can be consulted at:
doi.org/10.48550/arXiv.2112.1...

[@nicolasayobi]

Excellent explanation, the best I've seen

[@georgymarrero]

Very good presentation. Thank you.

[@user-lo7qh1ko3z]

Thank you. This helped me a lot :)

[@abhishekdhiman5719]

what is ground truth used for text guided inputs

[@deadpoetssociety7572]

Hi, Thanks for the presentation! Could you share your slides to me? Thanks!

[@jonatan01i]

They should ask participants later to draw a sketch from memory.
Or maybe scan other 20 participants' brain for the same images, regenerate from brain scans, and the participants have to guess which ones are their own.