the performance reported in your paper differs a lot from that in other papers

[shmily326]

Hi, thanks for your excellent work! But I'm confused about the big gap of Rank-1 accuracy of the same method(e.g Centerloss) w.r.t SCface reported in your paper between the one in this paper, which is 94.3% vs 31.71% when d=1, 81.8% vs 20.80% when d=2.6, and 36.3% vs 20.40% when d=4.2. Could you please help me out if I missed some key points?

Best regards.