Summary of Pseudo-ris: Distinctive Pseudo-supervision Generation For Referring Image Segmentation, by Seonghoon Yu et al.
Pseudo-RIS: Distinctive Pseudo-supervision Generation for Referring Image Segmentation
by Seonghoon Yu, Paul Hongsuck Seo, Jeany Son
First submitted to arxiv on: 10 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 The proposed framework automatically generates high-quality segmentation masks with referring expressions as pseudo supervisions for referring image segmentation (RIS). This allows the training of any supervised RIS methods without manual labeling. The framework incorporates existing segmentation and image captioning foundation models, leveraging their broad generalization capabilities. To generate distinctive captions, the framework proposes two-fold strategies: ‘distinctive caption sampling’, a new decoding method for the captioning model, and ‘distinctiveness-based text filtering’ to validate and filter out low-distinctiveness candidates. The generated pseudo supervisions significantly outperform weakly and zero-shot state-of-the-art methods on RIS benchmark datasets and surpass fully supervised methods in unseen domains. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new way to help computers understand images better is being developed. Right now, it’s hard to train computer models to recognize objects in pictures without lots of labeled data. But this new approach uses old models that can already do some tasks and adds special labels to help them learn even more. This makes the training process faster and cheaper. The method also includes a way to make sure the labels are clear and helpful, by giving the computer multiple options and letting it choose the best one. This approach is really good at recognizing objects in pictures and could be used to help computers do lots of other tasks too. |
Keywords
» Artificial intelligence » Generalization » Image captioning » Image segmentation » Supervised » Zero shot
