Summary of Evaluating the Robustness Of Off-road Autonomous Driving Segmentation Against Adversarial Attacks: a Dataset-centric Analysis, by Pankaj Deoli et al.
Evaluating the Robustness of Off-Road Autonomous Driving Segmentation against Adversarial Attacks: A Dataset-Centric analysis
by Pankaj Deoli, Rohit Kumar, Axel Vierling, Karsten Berns
First submitted to arxiv on: 3 Feb 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 The paper explores the vulnerability of semantic segmentation models in off-road autonomous driving to adversarial input perturbations. Despite good performance, state-of-the-art classifiers are often susceptible to small perturbations, leading to inaccurate predictions with high confidence. The study aims to address this gap by examining the impact of non-robust features in off-road datasets and comparing the effects of adversarial attacks on different segmentation network architectures. To achieve this, a robust dataset is created consisting only of robust features and training networks on this robustified dataset. The paper presents both qualitative and quantitative analysis of findings, which have important implications for improving the robustness of machine learning models in off-road autonomous driving applications. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The study looks at how good self-driving cars are at recognizing what’s around them when someone tries to trick them with fake pictures. Even though they’re really good, they can get fooled by tiny changes. The researchers want to know why this happens and how to make the self-driving cars better at recognizing things even when they’re trying to trick them. To do this, they made a special dataset that only has trustworthy information and trained the self-driving car programs on it. They then tested these programs to see if they could still be fooled by fake pictures. |
Keywords
* Artificial intelligence * Machine learning * Semantic segmentation
