From db4cef5471431b21e362e694190b75a57af87b56 Mon Sep 17 00:00:00 2001
From: 1996scarlet <1996scarlet@gmail.com>
Date: Tue, 20 Oct 2020 22:06:31 +0800
Subject: [PATCH] update readme
---
Readme.md | 70 +++++++++++++++++++++++++++++++------------------------
1 file changed, 40 insertions(+), 30 deletions(-)
diff --git a/Readme.md b/Readme.md
index a6379f8..a439117 100644
--- a/Readme.md
+++ b/Readme.md
@@ -1,11 +1,12 @@
# OpenVtuber-虚拟爱抖露共享计划
-## Kizuna-Ai MMD demo : face capture via single RGB camera
+Kizuna-Ai MMD demo : face capture via single RGB camera
## Installation
+
### Requirements
* Python 3.5+
@@ -22,56 +23,65 @@ While not required, for optimal performance(especially for the detector) it is h
* `python3.7 ./PythonClient/vtuber_usb_camera.py --gpu -1`
## 人脸检测 （Face Detection）
-* [RetinaFace: Single-stage Dense Face Localisation in the Wild](https://arxiv.org/abs/1905.00641)
-* [RetinaFace (mxnet version)](https://github.com/deepinsight/insightface/tree/master/RetinaFace)
-
-RetinaFace is a practical single-stage [SOTA](http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html) face detector which is initially described in [arXiv technical report](https://arxiv.org/abs/1905.00641)
-
+[RetinaFace: Single-stage Dense Face Localisation in the Wild](https://openaccess.thecvf.com/content_CVPR_2020/html/Deng_RetinaFace_Single-Shot_Multi-Level_Face_Localisation_in_the_Wild_CVPR_2020_paper.html) of **CVPR 2020**, is a practical single-stage [SOTA](http://shuoyang1213.me/WIDERFACE/WiderFace_Results.html) face detector. It is highly recommended to read the official repo [RetinaFace (mxnet version)](https://github.com/deepinsight/insightface/tree/master/RetinaFace).
-
+However, since the detection target of the face capture system is in the middle-close range, there is no need for complex pyramid scaling. We designed and published [Faster RetinaFace](https://github.com/1996scarlet/faster-mobile-retinaface) to trade off between speed and accuracy, which can reach 500~1000 fps on normal laptops.
-## 头部姿态估计（Head Pose Estimation）
-* [head-pose-estimation](https://github.com/lincolnhard/head-pose-estimation)
+| Plan | Inference | Postprocess | Throughput Capacity (FPS)
+| --------|-----|--------|---------
+| 9750HQ+1660TI | 0.9ms | 1.5ms | 500~1000
+| Jetson-Nano | 4.6ms | 11.4ms | 80~200
## 特征点检测（Facial Landmarks Tracking）
+
The 2D pre-trained model is from the [deep-face-alignment](https://github.com/deepinx/deep-face-alignment) repository.
+
* Algorithm from [TPAMI 2019](https://arxiv.org/pdf/1808.04803.pdf)
* Training set is based on i-bug 300-W datasets. It's annotation is shown below:
-## 注视估计（Gaze Estimation)
+## 头部姿态估计（Head Pose Estimation）
-- [Laser Eye : Gaze Estimation via Deep Neural Networks](https://github.com/1996scarlet/Laser-Eye)
+* [head-pose-estimation](https://github.com/lincolnhard/head-pose-estimation)
-## MMD Loader
+## 注视估计（Gaze Estimation)
-- [Three.js Webgl Loader](https://threejs.org/examples/?q=MMD#webgl_loader_mmd)
+* [Laser Eye : Gaze Estimation via Deep Neural Networks](https://github.com/1996scarlet/Laser-Eye)
-## Live2D
+## MMD Loader
-- [插件版本](https://github.com/EYHN/hexo-helper-live2d)
-- [打包版本](https://github.com/galnetwen/Live2D)
+We apply [Three.js Webgl Loader](https://threejs.org/examples/?q=MMD#webgl_loader_mmd) to render MMD model on web pages.
-## Thanks
+## Special Thanks
-- [threejs.org](https://threejs.org/)
-- [kizunaai.com](http://kizunaai.com/)
+* [threejs.org](https://threejs.org/)
+* [kizunaai.com](http://kizunaai.com/)
## Citation
-```
+``` bibtex
+@misc{sun2020backbone,
+ title={A Backbone Replaceable Fine-tuning Network for Stable Face Alignment},
+ author={Xu Sun and Yingjie Guo and Shihong Xia},
+ year={2020},
+ eprint={2010.09501},
+ archivePrefix={arXiv},
+ primaryClass={cs.CV}
+}
+
@article{Bulat2018Hierarchical,
- title={Hierarchical binary CNNs for landmark localization with limited resources},
- author={Bulat, Adrian and Tzimiropoulos, Yorgos},
- journal={IEEE Transactions on Pattern Analysis & Machine Intelligence},
- year={2018},
+ title={Hierarchical binary CNNs for landmark localization with limited resources},
+ author={Bulat, Adrian and Tzimiropoulos, Yorgos},
+ journal={IEEE Transactions on Pattern Analysis & Machine Intelligence},
+ year={2018},
}
-
-@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}
+
+@InProceedings{Deng_2020_CVPR,
+ author = {Deng, Jiankang and Guo, Jia and Ververas, Evangelos and Kotsia, Irene and Zafeiriou, Stefanos},
+ title = {RetinaFace: Single-Shot Multi-Level Face Localisation in the Wild},
+ booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
+ month = {June},
+ year = {2020}
}
```