Lune

CVPR2024

MarkovGen: Structured Prediction for Efficient Text-to-Image Generation

Sadeep Jayasumana, Daniel Glasner, Srikumar Ramalingam, Andreas Veit, Ayan Chakrabarti, Sanjiv Kumar

DOI出版方

摘要

Fewer steps 1.5x faster Full Muse: All steps MarkovGen: Fewer steps + MRF 1.5x faster Figure 1. MarkovGen improves the speed and quality of token-based image generation models such as Muse, by reducing the number of sampling steps and replacing them with a light-weight Markov Random Field (MRF) model.