Summary of On Inherent Adversarial Robustness Of Active Vision Systems, by Amitangshu Mukherjee et al.
On Inherent Adversarial Robustness of Active Vision Systems
by Amitangshu Mukherjee, Timur Ibrayev, Kaushik Roy
First submitted to arxiv on: 29 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 The paper proposes incorporating active vision mechanisms into deep learning systems to improve their robustness against adversarial examples. It suggests that the human visual system’s ability to process different regions of an image with varying resolutions could be a key to making DNNs more resilient. The authors demonstrate the effectiveness of two active vision methods, GFNet and FALcon, in resisting state-of-the-art adversarial attacks, achieving 2-3 times greater robustness than standard convolutional networks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about making computer programs like Deep Neural Networks (DNNs) less vulnerable to fake information. This is important because DNNs are used for things like recognizing pictures and faces, but they can be tricked into doing the wrong thing if someone gives them special “noise” that looks like normal data. The researchers think this is happening because DNNs process all the information in an image equally, just like we look at everything in front of us. But humans don’t do that – our eyes move around and focus on different parts of what we’re looking at. If we could make computer programs work more like that, they might be less likely to fall for fake information. |
Keywords
» Artificial intelligence » Deep learning
