SkeleVision: Towards Adversarial Resiliency of Person Tracking with Multi-Task Learning (2022-04-02T00:00:00.000000Z)

TL;DR

This work studies the impact of multi-task learning (MTL) on the adversarial robustness of the widely used SiamRPN tracker, in the context of person tracking, and investigates the effect of jointly learning with semantically analogous tasks of persontracking and human keypoint detection.

Abstract

Person tracking using computer vision techniques has wide ranging applications such as autonomous driving, home security and sports analytics. However, the growing threat of adversarial attacks raises serious concerns regarding the security and reliability of such techniques. In this work, we study the impact of multi-task learning (MTL) on the adversarial robustness of the widely used SiamRPN tracker, in the context of person tracking. Specifically, we investigate the effect of jointly learning with semantically analogous tasks of person tracking and human keypoint detection. We conduct extensive experiments with more powerful adversarial attacks that can be physically realizable, demonstrating the practical value of our approach. Our empirical study with simulated as well as real-world datasets reveals that training with MTL consistently makes it harder to attack the SiamRPN tracker, compared to typically training only on the single task of person tracking.

Authors

Duen Horng Chau

7 papers

Nilaksh Das

1 papers

Shengyun Peng

1 papers

TL;DR

Abstract

Authors

References62 items

Visual Object Tracking on Multi-modal RGB-D Videos: A Review

Adversarial Robustness in Multi-Task Learning: Promises and Illusions

The Ninth Visual Object Tracking VOT2021 Challenge Results

DepthTrack: Unveiling the Power of RGBD Tracking

A real-time person tracking system based on SiamMask network for intelligent video surveillance

SiamMOT: Siamese Multi-Object Tracking

A Unified Multi-Scenario Attacking Network for Visual Object Tracking

Towards More Flexible and Accurate Object Tracking with Natural Language: Algorithms and Benchmark

IoU Attack: Towards Temporally Coherent Black-Box Adversarial Attack for Visual Object Tracking

Person Tracking and Reidentification for Multicamera Indoor Video Surveillance Systems

The Eighth Visual Object Tracking VOT2020 Challenge Results

Efficient Adversarial Attacks for Visual Object Tracking

Long-Term Action Dependence-Based Hierarchical Deep Association for Multi-Athlete Tracking in Sports Videos

Robust Tracking against Adversarial Attacks

DSiamMFT: An RGB-T fusion tracking method via dynamic Siamese networks using multi-layer feature fusion

Fooling Detection Alone is Not Enough: Adversarial Attack against Multiple Object Tracking

Deformable Siamese Attention Networks for Visual Object Tracking

Multi-Player Tracking for Multi-View Sports Videos with Improved K-Shortest Path Algorithm

Adversarial T-Shirt! Evading Person Detectors in a Physical World

The Seventh Visual Object Tracking VOT2019 Challenge Results

Adversarial Attacks and Defenses in Images, Graphs and Text: A Review

SiamFT: An RGB-Infrared Fusion Tracking Method via Fully Convolutional Siamese Networks

MultiDepth: Single-Image Depth Estimation via Multi-Task Regression and Classification

Multi-Person 3D Pose Estimation and Tracking in Sports

Robust Person Tracking Algorithm Based on Convolutional Neural Network for Indoor Video Surveillance Systems

Adversarial Examples Are Not Bugs, They Are Features

Physical Adversarial Textures That Fool Visual Object Tracking

Unsupervised Traffic Accident Detection in First-Person Videos

Learning Multi-Task Correlation Particle Filters for Visual Tracking

SiamRPN++: Evolution of Siamese Visual Tracking With Very Deep Networks

LaSOT: A High-Quality Benchmark for Large-Scale Single Object Tracking

Distractor-aware Siamese Networks for Visual Object Tracking

MultiPoseNet: Fast Multi-Person Pose Estimation using Pose Residual Network

Adversarial Robustness Toolbox v1.0.0

High Performance Visual Tracking with Siamese Region Proposal Network

Robust Physical-World Attacks on Deep Learning Visual Classification

ShapeShifter: Robust Physical Adversarial Attack on Faster R-CNN Object Detector

Adversarially Robust Generalization Requires More Data

First-Order Adversarial Vulnerability of Neural Networks and Input Dimension

Deep Defense: Training DNNs with Improved Adversarial Robustness

Future Person Localization in First-Person Videos

Long-Term On-board Prediction of People in Traffic Scenes Under Uncertainty

CARLA: An Open Urban Driving Simulator

Boosting Adversarial Attacks with Momentum

Learning Dynamic Siamese Network for Visual Object Tracking

A Survey on Multi-Task Learning

Multi-task Correlation Particle Filter for Robust Object Tracking

Towards Deep Learning Models Resistant to Adversarial Attacks

Ensemble Adversarial Training: Attacks and Defenses

Parseval Networks: Improving Robustness to Adversarial Examples

Adversarial Multi-task Learning for Text Classification

Fully-Convolutional Siamese Networks for Object Tracking

Learning to Track at 100 FPS with Deep Regression Networks

Object Tracking Benchmark

Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

Explaining and Harnessing Adversarial Examples

Microsoft COCO: Common Objects in Context

Intriguing properties of neural networks

Manifold Regularized Multitask Learning for Semi-Supervised Multilabel Image Classification

ImageNet classification with deep convolutional neural networks

Robust Visual Tracking via Structured Multi-Task Sparse Learning

Two Six Technologies: ARMORY

Field of Study