mirror of
https://github.com/deepinsight/insightface.git
synced 2026-05-19 07:27:52 +00:00
71 lines
2.5 KiB
Markdown
71 lines
2.5 KiB
Markdown
# RetinaFace Face Detector
|
|
|
|
## Introduction
|
|
|
|
RetinaFace is a practical single-stage face detector which is initially described in [arXiv technical report](https://arxiv.org/abs/1905.00641)
|
|
|
|
|
|
|
|
|
|
|
|
## Data
|
|
|
|
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:
|
|
|
|
```Shell
|
|
data/retinaface/
|
|
train/
|
|
images/
|
|
label.txt
|
|
val/
|
|
images/
|
|
label.txt
|
|
test/
|
|
images/
|
|
label.txt
|
|
```
|
|
|
|
## Training
|
|
|
|
Please check ``train.py`` for training.
|
|
1. Copy ``rcnn/sample_config.py`` to ``rcnn/config.py``
|
|
2. Download pretrained models and put them into ``model/``. TODO_LINK
|
|
3. Start training with ``CUDA_VISIBLE_DEVICES='0,1,2,3' python -u train.py --prefix ./model/retina --network resnet``. You may want to check the ``resnet`` network configuration in ``rcnn/config.py`` before starting, like pretrained model path, anchor setting and learning rate policy etc..
|
|
4. Basically we have two predefined network settings called ``resnet``(for medium and large size models) and ``mnet``(for lightweight models).
|
|
|
|
## Testing
|
|
|
|
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.
|
|
|
|
## References
|
|
|
|
```
|
|
@inproceedings{yang2016wider,
|
|
title = {WIDER FACE: A Face Detection Benchmark},
|
|
author = {Yang, Shuo and Luo, Ping and Loy, Chen Change and Tang, Xiaoou},
|
|
booktitle = {CVPR},
|
|
year = {2016}
|
|
}
|
|
|
|
@inproceedings{deng2019retinaface,
|
|
title={RetinaFace: Single-stage Dense Face Localisation in the Wild},
|
|
author={Deng, Jiankang and Guo, Jia and Yuxiang, Zhou and Jinke Yu and Irene Kotsia and Zafeiriou, Stefanos},
|
|
booktitle={arxiv},
|
|
year={2019}
|
|
}
|
|
```
|
|
|
|
|