Adapting DINOv3 via SimMIM and composite metric learning on U.S. IDs yields 99.83% Canadian layout accuracy and surfaces 276 fraud cases (222 missed by prior detectors) in 20k Canadian IDs via embedding analysis.
Gradual tuning: a better way of fine tuning the parameters of a deep neural network.arXiv preprint arXiv:1711.10177
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Layout-Aware Representation Learning for Open-Set ID Fraud Discovery
Adapting DINOv3 via SimMIM and composite metric learning on U.S. IDs yields 99.83% Canadian layout accuracy and surfaces 276 fraud cases (222 missed by prior detectors) in 20k Canadian IDs via embedding analysis.