Summary of Flow Priors For Linear Inverse Problems Via Iterative Corrupted Trajectory Matching, by Yasi Zhang and Peiyu Yu and Yaxuan Zhu and Yingshan Chang and Feng Gao and Ying Nian Wu and Oscar Leong
Flow Priors for Linear Inverse Problems via Iterative Corrupted Trajectory Matching
by Yasi Zhang, Peiyu Yu, Yaxuan Zhu, Yingshan Chang, Feng Gao, Ying Nian Wu, Oscar Leong
First submitted to arxiv on: 29 May 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Generative models based on flow matching have gained popularity for their simplicity and superior performance in high-resolution image synthesis. By leveraging the instantaneous change-of-variables formula, one can directly compute image likelihoods from a learned flow, making them attractive candidates as priors for downstream tasks such as inverse problems. The paper proposes an iterative algorithm to approximate the maximum-a-posteriori (MAP) estimator efficiently solve various linear inverse problems. The algorithm is mathematically justified by approximating the MAP objective with a sum of local MAP objectives. Gradient steps are performed sequentially to optimize these objectives. The approach is validated for super-resolution, deblurring, inpainting, and compressed sensing tasks, outperforming other methods based on flow matching. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper uses special computer models to help solve tricky problems like restoring blurry pictures or removing noise from audio recordings. These models are very good at guessing what the original picture or sound might look or sound like. The problem is that these models can be very slow and take a long time to work out their answers. To speed things up, the authors came up with a new way of using these models to get the answers they need quickly. They tested this method on several different kinds of problems and showed it works better than other methods. |
Keywords
» Artificial intelligence » Image synthesis » Super resolution
