A Dual-stage Prompt-driven Privacy-preserving Paradigm for Person Re-Identification
This work addresses privacy concerns in person re-identification by improving virtual data methods, offering a domain-specific incremental advance.
The paper tackles the challenge of poor domain generalization and complex construction in virtual datasets for person re-identification by proposing a dual-stage prompt-driven paradigm, which generates a large-scale virtual dataset (GenePerson with 130,519 images) and uses a prompt-driven disentanglement mechanism to achieve state-of-the-art generalization performance.
With growing concerns over data privacy, researchers have started using virtual data as an alternative to sensitive real-world images for training person re-identification (Re-ID) models. However, existing virtual datasets produced by game engines still face challenges such as complex construction and poor domain generalization, making them difficult to apply in real scenarios. To address these challenges, we propose a Dual-stage Prompt-driven Privacy-preserving Paradigm (DPPP). In the first stage, we generate rich prompts incorporating multi-dimensional attributes such as pedestrian appearance, illumination, and viewpoint that drive the diffusion model to synthesize diverse data end-to-end, building a large-scale virtual dataset named GenePerson with 130,519 images of 6,641 identities. In the second stage, we propose a Prompt-driven Disentanglement Mechanism (PDM) to learn domain-invariant generalization features. With the aid of contrastive learning, we employ two textual inversion networks to map images into pseudo-words representing style and content, respectively, thereby constructing style-disentangled content prompts to guide the model in learning domain-invariant content features at the image level. Experiments demonstrate that models trained on GenePerson with PDM achieve state-of-the-art generalization performance, surpassing those on popular real and virtual Re-ID datasets.