• Corpus ID: 85501044

Metric Attack and Defense for Person Re-identification

@article{Bai2019MetricAA,
  title={Metric Attack and Defense for Person Re-identification},
  author={Song Bai and Yingwei Li and Yuyin Zhou and Qizhu Li and Philip H. S. Torr},
  journal={arXiv: Computer Vision and Pattern Recognition},
  year={2019}
}
Person re-identification (re-ID) has attracted much attention recently due to its great importance in video surveillance. [] Key Result At last, by benchmarking various adversarial settings, we expect that our work can facilitate the development of adversarial attack and defense in metric-based applications.

Figures and Tables from this paper

Vulnerability of Person Re-Identification Models to Metric Adversarial Attacks
TLDR
This article investigates different possible attacks depending on the number and type of guides available and proposes an efficient extension of adversarial training protocol adapted to metric learning as a defense that increases the robustness of re-ID models.
An Effective Adversarial Attack on Person Re-Identification in Video Surveillance via Dispersion Reduction
TLDR
This paper presents an effective and generalizable attack model that generates adversarial images of people, and results in very significant drop in the performance of the existing state-of-the-art person re-identification models.
Generative Metric Learning for Adversarially Robust Open-World Person Re-Identification
  • Deyin Liu, L. Wu, Bennamoun
  • Computer Science
    ACM Transactions on Multimedia Computing, Communications, and Applications
  • 2022
TLDR
This work improves the robustness of open-world re-ID models by proposing a generative metric learning approach to generate adversarial examples that are regularized to produce robust distance metric.
A Person Re-identification Data Augmentation Method with Adversarial Defense Effect
TLDR
The proposed ReID multi-modal data augmentation method performs well on multiple datasets, and successfully defends the attack of MS-SSIM proposed by CVPR2020 against ReID, and increases the accuracy by 467 times.
Learning to Attack Real-World Models for Person Re-identification via Virtual-Guided Meta-Learning
TLDR
This study argues that learning powerful attackers with high universality that works well on unseen domains is an important step in pro-moting the robustness of re-ID systems and proposes a novel universal attack algorithm called “MetaAt-tack” for person re- ID.
Robust Person Re-identification with Multi-Modal Joint Defence
TLDR
The Person Re-identification (ReID) system based on metric learning has been proved to inherit the vulnerability of deep neural networks, which are easy to be fooled by adversarail metric attacks, so this work proposes targeted methods for metric attacks and defence methods.
Transferable, Controllable, and Inconspicuous Adversarial Attacks on Person Re-identification With Deep Mis-Ranking
TLDR
This work examines the insecurity of current best-performing ReID models by proposing a learning-to-mis-rank formulation to perturb the ranking of the system output and performs a back-box attack by developing a novel multi-stage network architecture that pyramids the features of different levels to extract general and transferable features for the adversarial perturbations.
On the Robustness of Domain Adaption to Adversarial Attacks
TLDR
This work discusses the robustness of unsupervised domain adaption against adversarial attacking for the first time, benchmarking various settings of adversarial attack and defense in domainAdaption, and proposes a cross domain attack method based on pseudo label.
Detecting adversarial attacks on audio-visual speech recognition using deep learning method
TLDR
Experimental results have indicated that the proposed strategy is a powerful method to identify the adversarial attacks compared to Supervised Kernel Machines, Combined Neural Network, and Band Feature Selection methods.
Regional Homogeneity: Towards Learning Transferable Universal Adversarial Perturbations Against Defenses
TLDR
It is shown that a simple universal perturbation can fool a series of state-of-the-art defenses, and it is verified that regionally homogeneous perturbations can well transfer across different vision tasks.
...
...

References

SHOWING 1-10 OF 60 REFERENCES
Adversarial Binary Coding for Efficient Person Re-Identification
TLDR
The proposed Adversarial Binary Coding (ABC) implicitly fits the feature distribution to the expected binary one by optimizing the Wasserstein distance and seamlessly embeds the ABC into a similarity measuring deep neural network to enhance the semantic discriminability of binary codes.
Adversarial Attribute-Image Person Re-identification
TLDR
This work conducted extensive experiments on three attribute datasets and demonstrated that the regularized adversarial modelling is so far the most effective method for the attribute-image cross-modality person Re-ID problem.
Adversarially Occluded Samples for Person Re-identification
TLDR
Adversarially Occluded Samples are proposed to augment the variation of training data by introducing special samples that resemble real-scene occlusions and help the model discover new discriminative clues on the body and generalize much better at test time.
Pose-Normalized Image Generation for Person Re-identification
TLDR
This work proposes a novel deep person image generation model for synthesizing realistic person images conditional on the pose based on a generative adversarial network designed specifically for pose normalization in re-id, thus termed pose-normalization GAN (PN-GAN).
A Systematic Evaluation and Benchmark for Person Re-Identification: Features, Metrics, and Datasets
TLDR
This work presents an extensive review and performance evaluation of single- and multi-shot re-id algorithms and incorporates the most recent advances in both feature extraction and metric learning.
Person Transfer GAN to Bridge Domain Gap for Person Re-identification
TLDR
A Person Transfer Generative Adversarial Network (PTGAN) is proposed to relieve the expensive costs of annotating new training samples and comprehensive experiments show that the domain gap could be substantially narrowed-down by the PTGAN.
Adversarial Examples for Semantic Segmentation and Object Detection
TLDR
This paper proposes a novel algorithm named Dense Adversary Generation (DAG), which applies to the state-of-the-art networks for segmentation and detection, and finds that the adversarial perturbations can be transferred across networks with different training data, based on different architectures, and even for different recognition tasks.
Pose Transferrable Person Re-identification
TLDR
A pose-transferrable person ReID framework which utilizes posetransferred sample augmentations (i.e., with ID supervision) to enhance ReID model training, and achieves great performance improvement, and outperforms most state-of-the-art methods without elaborate designing the ReIDs.
DeepReID: Deep Filter Pairing Neural Network for Person Re-identification
TLDR
A novel filter pairing neural network (FPNN) to jointly handle misalignment, photometric and geometric transforms, occlusions and background clutter is proposed and significantly outperforms state-of-the-art methods on this dataset.
Harmonious Attention Network for Person Re-identification
  • Wei Li, Xiatian Zhu, S. Gong
  • Computer Science
    2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition
  • 2018
TLDR
A novel Harmonious Attention CNN (HA-CNN) model is formulated for joint learning of soft pixel attention and hard regional attention along with simultaneous optimisation of feature representations, dedicated to optimise person re-id in uncontrolled (misaligned) images.
...
...