Self-supervised Adversarial Training of Monocular Depth Estimation against Physical-World Attacks

Zhiyuan Cheng, Cheng Han, James Liang, Qifan Wang, Xiangyu Zhang +1 more
6/9/2024
cs.CV

No official code found. Best implementation: Bob-cheng/DepthModelHardening (11 stars, MIT, last commit 2023-02-08).

Abstract

Monocular Depth Estimation (MDE) plays a vital role in applications such as autonomous driving. However, various attacks target MDE models, with physical attacks posing significant threats to system security. Traditional adversarial training methods, which require ground-truth labels, are not directly applicable to MDE models that lack ground-truth depth. Some self-supervised model hardening techniques (e.g., contrastive learning) overlook the domain knowledge of MDE, resulting in suboptimal performance. In this work, we introduce a novel self-supervised adversarial training approach for MDE models, leveraging view synthesis without the need for ground-truth depth. We enhance adversarial robustness against real-world attacks by incorporating L_0-norm-bounded perturbation during training. We evaluate our method against supervised learning-based and contrastive learning-based approaches specifically designed for MDE. Our experiments with two representative MDE networks demonstrate improved robustness against various adversarial attacks, with minimal impact on benign performance.

View on arXivView PDF

Code Implementations(1)

Bob-cheng/DepthModelHardeningCommunity100%
MIT

Official PyTorch implementation of our paper "Adversarial Training of Self-supervised Monocular Depth Estimation against Physical-World Attacks" accepted at ICLR23 (Spotlight).

113Shell, PythonFeb 8, 20233 years agoMIT

Similar Papers

Adversarial Training of Self-supervised Monocular Depth Estimation against Physical-World Attacks

98% match1 repo112026

Self-Supervised Adversarial Training via Diverse Augmented Queries and Self-Supervised Double Perturbation

89% match1 repo2026

Adversarial Robustness of Discriminative Self-Supervised Learning in Vision

87% match2025

Pulling Back the Curtain: Unsupervised Adversarial Detection via Contrastive Auxiliary Networks

86% match2025

Attack Assessment and Augmented Identity Recognition for Human Skeleton Data

86% match2026

Improving Domain Generalization in Self-supervised Monocular Depth Estimation via Stabilized Adversarial Training

86% match2026

Cite this paper

@article{cheng2024selfsupervised,
  title  = {Self-supervised Adversarial Training of Monocular Depth Estimation against Physical-World Attacks},
  author = {Zhiyuan Cheng and Cheng Han and James Liang and Qifan Wang and Xiangyu Zhang and Dongfang Liu},
  year   = {2024},
  eprint = {2406.05857},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2406.05857v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2406.05857.svg)](https://codeofpaper.com/paper/2406.05857)

Discussion