Face Verification is a complex problem and is evolving every year. There are many challenges one can face while creating the Face Verifcation System. To solve this problem there is a popular approach of One Shot learning.
WORK UNDER PROGRESS
Face.Verification.mp4
The is a network which has shared weights b/w all the data. There are multiple architectures to build this network. Trying with multiple networks the Resnet34 based architecture gives the good result. The training accuracy is about 85% and validation accuracy is about 79%.
The Triplet Loss function is used to achive the results.