S³FD: Single Shot Scale-invariant Face Detector

Introduction

S³FD is a real-time face detector, which performs superiorly on various scales of faces with a single deep neural network, especially for small faces. For more details, please refer to our arXiv paper. This repository intends to reproduce the results of the aforementioned paper.

Contents

1. Preparation
2. Evaluation
3. Training
4. Future Enhancements

Preparation

Follow the instructions here to download and install the proper Caffe version.

Evaluation

To test the pre-trained models, read the Test Instructions

Training

To train new models using SFD, read the Train Instructions

Enhancements

CFENet to enhance small object detection through feature fusion

• CFENet Paper
• CFENet Code

PyramidMobileNet to add greater spatial information to convolutions

• PyramidMobileNet Paper

DSFD uses a Feature Enhance Module with two shot pass

• DSFD Paper

SRN - Selective Refinement Network

• SRN Paper