VGGFace2-pytorch

VGGFace2-pytorch copied to clipboard

[Help] How to perform test/validation correctly on pre-trained models?

1

Hi, I notice that the training dataset (8631 classes in total) and the validation dataset (500 classes in total) seem to be disjoint, and according to the following code https://github.com/cydonia999/VGGFace2-pytorch/blob/c6e10f277b31b972c78fac68a40464a36a46a10d/utils.py#L33-L45...

wutongshenqiu

Can not find identity_meta.csv

4

It seems that the link your provide for identity_meta.csv is invalid and I would like to use VggFace2 as an extractor. Could you please send the meta file to [email protected]?

GaryGky

IOError: [Errno 2] No such file or directory: '/path/to/identity_meta.csv'

8

When I download the pre-train model and meet the error: IOError: [Errno 2] No such file or directory: '/path/to/identity_meta.csv' Could you provide complete files? Thanks!

leoliu37

Feature Extraction on own pictures

4

Hi, thank you for posting this implementation. I wonder know is it possible that use the pre-trained model as a feature extractor and compare the distance on my own pictures?...

kevinisbest

Hello，i am not familiar with the difference between resnet50 model and resnet50-scratch model, could you please inform me. Thanks!

Watebear

Thanks for your demo, when I read your work in `demo.py`, one variable called `N_IDENTITY` defined as 8631, labeled the number of identities in VGGFace2 for which ResNet and SENet...

RuoyuChen10

Wrong number of classes in validation

4

Hi, I took the train/test CSV files from here [vggface2 meta](http://www.robots.ox.ac.uk/~vgg/data/vgg_face2/meta_infor.html) and when I try to test the model (or validate) it crashes in the loss criteria. I think it...

IdanAzuri

DENGZAIXU

resumable link to VGGface2 dataset

2

Hello Author, is it possible for you to share the database if you have downloaded it at your end. please share via Dropbox or any other method. I am not...

Sawatdatta

How you label the face images in Vggface2 dataset?

1

In the paper, I saw they used two pretrained classification model to get the information of head pose and age to label the images in Vggface2, now how you get...

Huixxi