A systematic study of five synthetic-data strategies for low-resolution face recognition on a compact backbone shows that simple interpolation augmentation beats learned super-resolution front-ends, and that synthetic benchmarks mis-rank strategies relative to real native-LR data.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Improving Low-Resolution Face Recognition under Limited Data: How Synthetic Data Generation Can Close the Domain Gap
A systematic study of five synthetic-data strategies for low-resolution face recognition on a compact backbone shows that simple interpolation augmentation beats learned super-resolution front-ends, and that synthetic benchmarks mis-rank strategies relative to real native-LR data.