search

leebebeto / Inter-Prototype

Official PyTorch implementation of "Improving Face Recognition with Large AgeGaps by Learning to Distinguish Children" (BMVC 2021)
BSD 3-Clause "New" or "Revised" License
18 stars 2 forks source link

Please share the weight file trained using LAG #3

Closed ma-siddiqui closed 2 years ago

ma-siddiqui commented 2 years ago

Hi,

I am interested to test and evaluate your trained model on LAG data set. Please share its weight file.

Thanks,

YeolJ00 commented 2 years ago

Hi there,

Thanks for your interest in our work.

Our models (and most publicly available face recognition models) are trained with a large-scale dataset (CASIA WebFace and MS1M) and evaluated on a different test-set (e.g., LFW, AgeDB, and LAG). Thus, if you intend to test our model on the LAG dataset, first download the model trained on the Casia WebFace or the MS1Mv2 and evaluate on the LAG test set.

To clarify once again, the LAG dataset or any other test sets are only used as test sets do not have a separate training split.

Thanks again!