AI system identifies people even when they change clothes
Researchers have developed a new approach to recognize individuals based on body shape and gait rather than clothing, solving a long-standing challenge in surveillance and security systems. The breakthrough could improve airport screening, law enforcement investigations, and loss prevention—while raising fresh privacy concerns for regulators and security teams.
Originaltitel: Disentangling identity from appearance: A semantic-hierarchical multi-level fusion framework for cloth-invariant person re-identification
<p>Cloth-changing person re-identification faces a fundamental challenge: identity-invariant and clothing-variant features are inherently entangled in low-level visual representations. Existing implicit disentanglement approaches lack explicit semantic supervision to define separation boundaries between identity and clothing attributes. Consequently, when individuals retain partial garments across observations, these methods produce spurious identity-clothing correlations that compromise recognition accuracy. We propose the Semantic-Hierarchical Disentanglement Network (SHD-Net), a framework that progressively decouples identity from clothing through multi-scale feature extraction and cross-architecture fusion. Our framework comprises three synergistic components. First, the Hierarchical Hybrid Feature Extraction Network employs dual CNN-ViT backbones with bidirectional cross-attention, fusing local textures and global semantics across shallow and deep layers to construct complementary multi-granularity identity representations. Second, the Semantic-Guided Disentanglement Module leverages semantic priors from large language models to explicitly supervise feature decomposition through a three-stage pipeline: initial separation, semantic-anchored refinement, and cross-attention recovery. Third, Confidence-Weighted Adaptive Fusion dynamically integrates purified features using prediction confidence as sample-specific weights, ensuring optimal feature combination. Extensive experiments demonstrate state-of-theart performance across multiple benchmarks.</p>