Loading Now

Summary of Attack on Scene Flow Using Point Clouds, by Haniyeh Ehsani Oskouie et al.


Attack on Scene Flow using Point Clouds

by Haniyeh Ehsani Oskouie, Mohammad-Shahram Moin, Shohreh Kasaei

First submitted to arxiv on: 21 Apr 2024

Categories

  • Main: Computer Vision and Pattern Recognition (cs.CV)
  • Secondary: Machine Learning (cs.LG); Multimedia (cs.MM)

     Abstract of paper      PDF of paper


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
Deep learning-based scene flow estimation using point clouds has made significant progress in applications like video analysis, action recognition, and navigation. However, the robustness of these techniques against adversarial attacks remains a concern, despite being a crucial aspect in many domains. To address this issue, this paper proposes an approach to introduce adversarial white-box attacks specifically designed for scene flow networks. Experimental results show that generated adversarial examples can cause up to 33.7% relative degradation in average end-point error on the KITTI and FlyingThings3D datasets. The study also highlights the significant impact of attacks targeting point clouds in one dimension or color channel, leading to increased vulnerability in optical flow networks. This research provides insights into the success and failure of these attacks, demonstrating the importance of robustness testing for scene flow estimation.
Low GrooveSquid.com (original content) Low Difficulty Summary
This paper is about making sure that computer vision systems are not fooled by fake data. These systems can be tricked into giving wrong answers if they’re given manipulated information. The researchers created special types of attacks to test how well these systems work. They found out that these attacks can cause a significant decrease in the accuracy of the systems, especially when they’re using point clouds to analyze scenes. This study is important because it shows us that we need to make sure our computer vision systems are robust and not easily fooled by fake data.

Keywords

» Artificial intelligence  » Deep learning  » Optical flow