Update README.md

RetinaFace/README.md

@@ -1,11 +1,31 @@
-### RetinaFace Face Detector
+# RetinaFace Face Detector
+
+## Introduction
+
+RetinaFace is a robust single stage face detector which initially described as an [arXiv technical report](https://arxiv.org/abs/1905.00641)
+
+
+## Training
+
+1. Download groundtruth labels from [baiducloud](https://pan.baidu.com/s/1Laby0EctfuJGgGMgRRgykA) or dropbox and organise the dataset directory under ``insightface/RetinaFace/`` as follows(images can be downloaded from WiderFace website directly):
+```Shell
+  data/retinaface/
+    train/
+      images/
+      label.txt
+    val/
+      images/
+      label.txt
+    test/
+      images/
+      label.txt
+```
+
+## Testing
 
 Please check ``test.py`` for model usage.
 
 Pretrained Model: [RetinaFace-R50](https://pan.baidu.com/s/1C6nKq122gJxRhb37vK0_LQ) is a medium size model with ResNet50 backbone. WiderFace validation mAP: Easy 96.5, Medium 95.6, Hard 90.4. It can output face bounding boxes and five landmarks in a single forward pass.
 
-Light weight and large models will be released soon.
 
-![example1](https://github.com/deepinsight/insightface/blob/master/resources/retina_R50_ex1.jpg)
 
-![example2](https://github.com/deepinsight/insightface/blob/master/resources/retina_R50_ex2.jpg)