2) Deep face recognition systems do not gener- alize well across benchmarks, due to the severe sampling biases in public datasets (as illustrated in Section 4).
確定! 回上一頁