New Machine Learning Method Improves Person Recognition Despite Clothing Changes
Researchers have developed Ortho-ReID, a machine learning technique that can identify individuals even when they change clothes by using geometric constraints and text descriptions to separate clothing features from identity features. The method uses a transformer-based system to create instance-adaptive subspaces that distinguish between what someone is wearing and who they are. This advancement could improve security and surveillance applications that need to track individuals across different appearances.
A new computer vision technique called Ortho-ReID addresses the challenge of clothes-changing person re-identification (CC-ReID) by explicitly modeling clothing features separately from identity features. Rather than relying solely on adversarial learning like previous methods, the approach uses geometric constraints and text embeddings from vision-language models to create low-rank clothing subspaces. The system includes a transformer-based component called Basis Maker that adapts these subspaces to individual images through cross-attention mechanisms, allowing it to extract identity information that remains invariant to clothing changes. The method was tested on multiple benchmark datasets (PRCC, Celeb-reID-light, LaST, and LTCC) and achieved state-of-the-art results on three of them, with improvements ranging from 3.5% to 5.9% in top-1 accuracy. The research was accepted to the ICML 2026 Workshop on CoLoRAI.
What's missing
The paper does not discuss potential privacy implications of clothes-changing person re-identification technology, nor does it address computational requirements or real-world deployment considerations for the method.
What different sources said
- arXiv cs.LGCenter
Learning Instance-Adaptive Low-Rank Orthogonal Subspaces for Clothes-Changing Person Re-Identification
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.