githuseyingur / computer_vision_demo_applications

Computer Vision Application Demos Notes & Model Training - Face Recognition with ArcFace - The Best Success Rate (ArcFace)

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

COMPUTER VISION DEMO APPLICATIONS

This repo contains notes from the 'Bilgisayarlı Görü Uygulama Alanları(Computer Vision Applications)' course on BTK Akademi.

1. FACE RECOGNITION WITH ARCFACE

  • Face Recognition Systems & Model Training
  • Only a part of the MS1M dataset was used : DATASET & FILES

SOTA : the State Of The Art

  • Working with the complete MS1M dataset
  • conf.mode = 'ir'
  • conf.depth = '100'
  • conf.total_epoch = 20
  • conf.milestones = [12,16,18]

lfw = gives an accuracy of 99.83%. (is the best rate ever obtained.)
It takes 5 days with two v100(32GB).

About

Computer Vision Application Demos Notes & Model Training - Face Recognition with ArcFace - The Best Success Rate (ArcFace)


Languages

Language:Jupyter Notebook 100.0%