Scale Attention for Learning Deep Face Representation: A Study Against Visual Scale Variation | Read Paper on Bytez