To this end, we propose multiplicative angular margin (MAM) loss to learn angularly discriminative features for each identity.
確定! 回上一頁