ZF/insightface
1
0
Fork 0
mirror of https://github.com/deepinsight/insightface.git synced 2026-05-18 14:55:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
JiankangDeng
2019-05-04 11:53:59 +01:00
committed by GitHub
parent 3580f341a5
commit 1bc1b33ca6
1 changed files with 20 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
RetinaFace/README.md
View File

@@ -2,14 +2,18 @@
## Introduction
RetinaFace is a robust single stage face detector which initially described as an [arXiv technical report](https://arxiv.org/abs/1905.00641)
RetinaFace is a practical single-stage face detector which is initially described in [arXiv technical report](https://arxiv.org/abs/1905.00641)
![demoimg](https://github.com/deepinsight/insightface/blob/master/resources/11513D05.jpg)
## Data
## Training
1. Download our annotations (face bounding boxes & five facial landmarks) from [baiducloud](https://pan.baidu.com/s/1Laby0EctfuJGgGMgRRgykA) or [dropbox](https://www.dropbox.com/s/7j70r3eeepe4r2g/retinaface_gt_v1.1.zip?dl=0)
2. Download the [WIDERFACE](http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html) dataset.
3. Organise the dataset directory under ``insightface/RetinaFace/`` as follows:
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/
@@ -23,11 +27,22 @@ RetinaFace is a robust single stage face detector which initially described as a
label.txt
```
## Training
Please check ``train.py`` for training.
## Testing
Please check ``test.py`` for model usage.
Please check ``test.py`` for testing.
## Models
Pretrained Model: RetinaFace-R50 ([baiducloud](https://pan.baidu.com/s/1C6nKq122gJxRhb37vK0_LQ) or [dropbox](https://www.dropbox.com/s/53ftnlarhyrpkg2/retinaface-R50.zip?dl=0)) is a medium size model with ResNet50 backbone.
It can output face bounding boxes and five facial landmarks in a single forward pass.
WiderFace validation mAP: Easy 96.5, Medium 95.6, Hard 90.4.
To avoid the confliction with the WiderFace Challenge (ICCV 2019), we postpone the release time of our best model.
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.