Abstract:The receptive fields of deep learning models determine the most significant regions of the input data for providing correct decisions. Up to now, the primary way to learn such receptive fields is to train the models upon masked data, which helps the networks to ignore any unwanted regions, but also has two major drawbacks: 1) it yields edge-sensitive decision processes; and 2) it augments considerably the computational cost of the inference phase. Having theses weaknesses in mind, this paper describes a solution for implicitly enhancing the inference of the networks' receptive fields, by creating synthetic learning data composed of interchanged segments considered apriori important or irrelevant for the network decision. In practice, we use a segmentation module to distinguish between the foreground (important) versus background (irrelevant) parts of each learning instance, and randomly swap segments between image pairs, while keeping the class label exclusively consistent with the label of the segments deemed important. This strategy typically drives the networks to interpret that the identity and clutter descriptions are not correlated. Moreover, the proposed solution has other interesting properties: 1) it is parameter-learning-free; 2) it fully preserves the label information; and 3) it is compatible with the data augmentation techniques typically used. In our empirical evaluation, we considered the person re-identification problem, and the well known RAP, Market1501 and MSMT-V2 datasets for two different settings (upper-body and full-body), having observed highly competitive results over the state-of-the-art. Under a reproducible research paradigm, both the code and the empirical evaluation protocol are available at <a href="https://github.com/Ehsan-Yaghoubi/reid-strong-baseline">https://github.com/Ehsan-Yaghoubi/reid-strong-baseline</a>.

SecureReID: Privacy-Preserving Anonymization for Person Re-Identification

Adversarial Recovery Network for Low-Light Person Re-Identification.

Learnable Privacy-Preserving Anonymization for Pedestrian Images

RETRACTED CHAPTER: Person Re-identification Based on Transform Algorithm

Person Re-identification Network Based on Multi-Level Feature Fusion

Towards Privacy-Preserving Person Re-identification via Person Identify Shift

Person Re-Identification without Identification via Event Anonymization

Secure Dual Network for Reversible Facial Image Anonymization Through the Latent Space Manipulation

Personalized and Invertible Face De-identification by Disentangled Identity Information Manipulation

Re-ID-leak: Membership Inference Attacks Against Person Re-identification

Joint Reconstruction and Deidentification for Mobile Identity Anonymization

Self-Erasing Network for Person Re-Identification

Occluded Person Re-Identification with Single-scale Global Representations

Learning Sparse and Identity-Preserved Hidden Attributes for Person Re-Identification

PixelFade: Privacy-preserving Person Re-identification with Noise-guided Progressive Replacement

Privacy-Preserving Adaptive Re-Identification without Image Transfer

Frustratingly Easy Person Re-Identification: Generalizing Person Re-ID in Practice

Image-Level Adaptive Adversarial Ranking for Person Re-Identification

Determining the Robustness of Privacy Enhancing DeID Against the ReID Adversary: An Experimental Study

Data-Driven but Privacy-Conscious: Pedestrian Dataset De-identification via Full-Body Person Synthesis

Person re-identification: Implicitly defining the receptive fields of deep learning classification frameworks