Summary of Partial Wasserstein Adversarial Network For Non-rigid Point Set Registration, by Zi-ming Wang et al.
Partial Wasserstein Adversarial Network for Non-rigid Point Set Registration
by Zi-Ming Wang, Nan Xue, Ling Lei, Gui-Song Xia
First submitted to arxiv on: 4 Mar 2022
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: None
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 This paper addresses the challenge of registering two point sets, a problem that’s difficult due to outliers, unknown deformations, and large dataset sizes. To overcome these issues, the authors formulate the registration problem as a partial distribution matching (PDM) task in a metric space. They propose a scalable PDM algorithm using the efficient partial Wasserstein-1 discrepancy, which involves deriving the Kantorovich-Rubinstein duality for this discrepancy. This is used to train a neural network-based partial Wasserstein adversarial network (PWAN), which can approximate the PW discrepancy and minimize it through gradient descent. The authors also incorporate an efficient coherence regularizer to avoid unrealistic deformations. Experiments show that PWAN is robust, scalable, and outperforms state-of-the-art methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us better match two sets of points by dealing with things like noise and shape changes. To do this, the researchers turn the matching problem into a different kind of math problem called partial distribution matching. They then create an algorithm that can handle large amounts of data and is good at ignoring noise. The algorithm also includes a special trick to make sure the shapes match in a way that looks realistic. The authors tested their method on some real-world problems and found it worked better than other methods. |
Keywords
* Artificial intelligence * Gradient descent * Neural network
