Synthetic data generation closes low-resolution face recognition domain gap

The paper targets surveillance face recognition where faces fall below the standard 112×112 input size, proposing synthetic data generation to bridge the domain gap when labeled high-resolution data is scarce.