# pytorch-QualityFace

**Repository Path**: siriusdemon/pytorch-QualityFace

## Basic Information

- **Project Name**: pytorch-QualityFace
- **Description**: Quality Face model which decides how suitable of an input face for face recognition system
- **Primary Language**: Python
- **License**: MIT
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 1
- **Forks**: 1
- **Created**: 2020-05-15
- **Last Updated**: 2021-12-27

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# pytorch-QualityFace

[English version](./README.md)

人脸质量模型，可用于评估一张图片是否适合用作人脸识别系统的输入。

请注意，`qualityface`仅是一个辅助模型，其输入必须刚好是 **_128x128x3_**。

## 快速开始

+ Python3.6 (test in 3.6, 3.7)
+ Pytorch (test in 1.3.1, 1.4.0)
+ opencv4 (test in 4.1.1)
+ pillow (test in 6.2.0)
+ numpy
+ siriusbackbone

install from PyPI:
```sh
pip install qualityface
```

## 用法

```sh
import qualityface
path = 'path/to/your/img'
score = qualityface.estimate(path)
print(score)
```

## 结果
<img src="test/crop1.jpeg">
<img src="test/crop2.jpeg">
<img src="test/crop3.jpeg">
<img src="test/crop4.jpeg">
<img src="test/crop5.jpeg">

```sh
Test: crop1.jpeg, score: 0.78
Test: crop2.jpeg, score: 0.7
Test: crop3.jpeg, score: 0.74
Test: crop4.jpeg, score: 0.92
Test: crop5.jpeg, score: 0.93
```

这个模型对于1）遮挡、2）非正脸、3）模糊、4）阴影、5）正常 都有比较好的判别度。

### 许可证

This code is distributed under the [MIT license](LICENSE).


### 祈愿

愿所有见闻忆触此仓库者，都能够获得暂时的快乐和永恒不变的快乐。