QiaoXie / Tensorflow_RetinaFace

RetinaFace: Single-stage Dense Face Localisation in the Wild

RetinaFace: Single-stage Dense Face Localisation in the Wild

# Difference from Paper

1. standard OHEM -> Focal Loss
2. Smooth L1 Loss -> GIoU
3. Momentum Optimizer -> Adam Optimizer

# Reference

• Focal Loss for Dense Object Detection [Paper]
• RetinaFace: Single-stage Dense Face Localisation in the Wild [Paper]
• Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regression [Paper] [Code]