-
Notifications
You must be signed in to change notification settings - Fork 5.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Ms1m/asian-celeb/deepglint dataset relationship ? #789
Comments
They have gone through different cleaning procedures I guess. i.e. those two datasets are two different versions of the the same datasetm, cleaned differently. There is another version called C-MS-Celeb which you can also use. it has 94K ids and 6.4M images |
@SueeH the deepglint includes Asian-celeb which was listed on website http://trillionpairs.deepglint.com/data, i plan to merge glint and ms1m-arcface datasets, but i can not map the label in ms1m-arc with glint's ms1m-v1c part. what should i do? @Coderx7 |
You may use face recognition to find same id. But I think use one dataset is fine, most of the id is same. |
@SueeH ok, thanks |
|
hi,Do you have any information about the Asian-celeb dataset?(^▽^) |
https://github.com/deepinsight/insightface/tree/master/recognition/_datasets_ |
Thank you. I'll go see now.(^▽^) |
I have read some issues, and found some discuss about dataset.
ms1m-v1 = ms1m-ibug, include 85k IDS/3.8M images
ms1m-v2 = ms1m-arc=emore include 85K IDS/5.8M images
I found in arc-loss paper, ms1m-ibug is used,while the pretrained model in open model zoo, ms1m-arc is used.
But sitll have some confusion:
this two datasets have same ids, but why the second one has 2M images more? someone said data augment is used? if really, augment types?
Also DeepGlint(in dataset zoo 181K IDS/6.75 images) is same with ori deepglint website(http://trillionpairs.deepglint.com/data)? or any other operations?
Asian-celeb(94K IDS/2.8M images) has relationship with deepglint?
@nttstar
The text was updated successfully, but these errors were encountered: