Faceanti-spoofingissignificanttothesecurityoffacerecognitionsystems.Previousworksondepthsuper-visedlearninghaveprovedtheeffectivenessforfaceanti-spoofing.Nevertheless,theyonlyconsideredthedepthasanauxiliarysupervisioninthesingleframe.Differentfromthesemethods,wedevelopanewmethodtoest