Open GuoGuoYao opened 5 years ago
I used MS-Celeb-1M dataset to train the moel, Do you have any different parameters on this data set?
The iteration of ms-celeb-1m training is larger than the casia-webface training one. The other parameters are the same. Note: The ms-celeb-1m used in our training is not the clean one but the original one.
I used MS-Celeb-1M dataset to train the moel, Do you have any different parameters on this data set?