Summary of Adversarial Testing For Visual Grounding Via Image-aware Property Reduction, by Zhiyuan Chang et al.
Adversarial Testing for Visual Grounding via Image-Aware Property Reduction
by Zhiyuan Chang, Mingyang Li, Junjie Wang, Cheng Li, Boyu Wu, Fanjiang Xu, Qing Wang
First submitted to arxiv on: 2 Mar 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 This paper proposes a novel approach to evaluate Visual Grounding (VG) models, which locate objects in images through natural language expressions. The authors address the limitations of existing adversarial testing techniques that fail to exploit the correlation between visual and textual information. They introduce PEELING, a text perturbation method that reduces property-related information while ensuring the reduced expression uniquely describes the original object. This approach is evaluated on the OFA-VG model using three datasets, achieving a MultiModal Impact score of 21.4% and outperforming state-of-the-art baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you’re trying to teach a computer to recognize objects in pictures based on what people say about them. It’s a tricky task! The problem is that most testing methods only look at one piece of information, either the picture or the words, without considering how they relate to each other. This can lead to inaccurate results. To fix this, researchers developed a new method called PEELING. It works by taking a sentence about an object and reducing it to a simpler form that still describes the same thing. They test this approach on a popular model for recognizing objects in pictures and find that it performs better than other methods. |
Keywords
» Artificial intelligence » Grounding
