Swapping Faces, Saving Features: A Dual-Purpose Pipeline for Pedestrian Privacy in ITS

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Roba H. Farouk, Catherine M. Elias
• Venue: arXiv: cs.CV
• Published: 2026-07-09
• Collected: 2026-07-10
• URL: http://arxiv.org/abs/2607.08402v1
• PDF: https://arxiv.org/pdf/2607.08402v1
• DOI: UNKNOWN

Plain-Language Brief

The paper focuses on creating a five-stage pipeline for protecting pedestrians' privacy in AI models used by autonomous vehicles (AVs) while maintaining the image attributes needed to train the models. The proposed solution uses the Roop face-swapping model, which outperforms Ghost-v2 in preserving both privacy and usability of the images.

What the Document Says

• The work aims to implement a five-stage pipeline for preserving pedestrian privacy in AI models for AVs.
• The pipeline uses face swapping to conceal identities while preserving essential facial attributes.
• The pipeline is tailored for the Egy-DRiVeS dataset.
• Two face-swapping models, Roop and Ghost-v2, are evaluated.
• Roop outperforms Ghost-v2 in various aspects, but specific details are not provided.

Key Concepts Mentioned

• Autonomous vehicles
• Intelligent Transportation System (ITS)
• Pedestrian privacy
• Face swapping
• Egy-DRiVeS dataset

Claims Made by the Document

• The proposed pipeline preserves pedestrian privacy while maintaining image usability.
• Roop outperforms Ghost-v2 in various aspects.

Evidence Used

• The text states that the work aims to implement a pipeline for preserving pedestrian privacy and maintains the essential facial attributes.
• It is mentioned that Roop outperforms Ghost-v2, but no specific details are provided.

Methods / Approach

• Implementation of a five-stage pipeline for face swapping.
• Evaluation of Roop and Ghost-v2 face-swapping models.

Limitations Stated by the Authors

• The text does not provide specific details about how Roop outperforms Ghost-v2.

Limitations Noted by the Agent

Agent interpretation — not from the source.

• The text does not specify the metrics used to evaluate the performance of Roop and Ghost-v2.
• The paper does not discuss potential challenges or limitations of the proposed solution beyond those mentioned by the authors.

Open Questions

• What specific aspects does Roop outperform Ghost-v2 in?
• How is the performance of the proposed pipeline evaluated?

Notes for Future Q&A

May include interpretation. Labeled where it does.

• The text suggests that Roop outperforms Ghost-v2, but it's unclear in what specific aspects.
• It's unknown how the performance of the proposed pipeline is being evaluated.

Raw Abstract / Excerpt

Large-scale and diverse datasets are needed to train AI models to take real-time decisions for autonomous vehicles (AVs), an intelligent transportation system (ITS) application. Pedestrian intention and trajectory prediction are critical models used in AVs, requiring datasets involving diverse pedestrian images. Unrestricted access to these datasets imposes serious security risks, like identity theft and pedestrian tracking. The challenge is to apply privacy preservation procedures while maintaining the image attributes needed to train the models. Existing privacy methods may preserve the pedestrian's privacy, but degrade the image usability, which hinders the models' effectiveness. This work's focus is to implement a five-stage pipeline to protect pedestrians' privacy through face swapping while keeping the essential facial attributes intact. It should be tailored to satisfy the privacy needs of the Egy-DRiVeS dataset. Moreover, Roop and Ghost-v2 face-swapping models are evaluated. Provenly, Roop outperforms Ghost-v2 in various aspects, as will be discussed. Consequently, Roop is the face-swapping model to be used in the pipeline to strike the balance between pedestrian privacy via identity concealment and data usability via facial attribute preservation.