Adversarial Attacks on Monocular Pose Estimation

TL;DR

This paper demonstrates how imperceptible adversarial perturbations can significantly disrupt monocular pose estimation, revealing vulnerabilities and transferability issues in current deep learning models used for 3D scene understanding.

Contribution

It provides a systematic analysis of adversarial attacks on monocular pose estimation, including their effects, transferability, and implications for network robustness.

Findings

Adversarial perturbations cause significant errors in pose predictions.

Perturbations transfer across different network architectures.

Adversarial attacks increase trajectory drift and geometric errors.

Abstract

Advances in deep learning have resulted in steady progress in computer vision with improved accuracy on tasks such as object detection and semantic segmentation. Nevertheless, deep neural networks are vulnerable to adversarial attacks, thus presenting a challenge in reliable deployment. Two of the prominent tasks in 3D scene-understanding for robotics and advanced drive assistance systems are monocular depth and pose estimation, often learned together in an unsupervised manner. While studies evaluating the impact of adversarial attacks on monocular depth estimation exist, a systematic demonstration and analysis of adversarial perturbations against pose estimation are lacking. We show how additive imperceptible perturbations can not only change predictions to increase the trajectory drift but also catastrophically alter its geometry. We also study the relation between adversarial perturbations targeting monocular depth and pose estimation networks, as well as the transferability of perturbations to other networks with different architectures and losses. Our experiments show how the generated perturbations lead to notable errors in relative rotation and translation predictions and elucidate vulnerabilities of the networks.