search

deepinsight / insightface

State-of-the-art 2D and 3D Face Analysis Project
https://insightface.ai
23.47k stars 5.42k forks source link

glint360k dataset overlap with other datasets. #1317

Open zwacke opened 4 years ago

zwacke commented 4 years ago

Hi,

first of all thank you for this huge dataset and congrats on the PartialFC paper, both look really promising.

In the paper you mention that the dataset consists of a cleaning of the Celeb500k data and some other publicly available datasets. I have some questions that you maybe can help me out with.

1) Do you mind sharing which other datasets are included? I would like to know whether there is overlap with like ms1m, vgg, asian celeb, so I avoid mixing or testing on false assumptions. 2) Do you have any info for the mapping from the original datasets' label/file ids to the label/file ids in glint360k? I would like to reallign the original source images to a different crop and format, but as far as I can tell in glint360k the label names are just enumeration indices.

Thanks so much in advance!

HaohaoNJU commented 3 years ago

I'm also interested with your Q1, I suppose there must be overlap with popular datasets like VGG,ms1s,casia, and so on !

lmtri1998 commented 3 years ago

Any update on this?