# Age-and-Gender

**Repository Path**: chenyang918/Age-and-Gender

## Basic Information

- **Project Name**: Age-and-Gender
- **Description**: 同时识别年龄与性别
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 1
- **Created**: 2020-08-30
- **Last Updated**: 2021-09-02

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# 同时识别年龄与性别
基于PyTorch 实现多任务学习，在同时识别年龄与性别。


## 数据集

[IMDB-WIKI](https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/) 数据集，460,723张图片。

![image](https://github.com/foamliu/Joint-Estimation-of-Age-and-Gender/raw/master/images/imdb-wiki-teaser.png)

这里为简洁只使用 IMDB 数据集。

### 年龄分布：

460723张照片为平衡每个年龄最多只保存5000张，清洗后得到163065张，按年龄分布作图：

![image](https://github.com/foamliu/Joint-Estimation-of-Age-and-Gender/raw/master/images/age.png)

## 用法

### 数据预处理
提取163065张图片：
```bash
$ python pre_process.py
```

### 训练
```bash
$ python train.py
```

### 准确度比较

#|图片大小|网络|性别准确度(验证集)|年龄MAE(验证集)|年龄损失函数|批量大小|Loss|年龄权重|
|---|---|---|---|---|---|---|---|---|
|1|112x112|ResNet-18|90.756%|5.710|回归(L1Loss)|32|0.9757|0.1|
|2|224x224|ResNet-18|90.887%|5.694|回归(L1Loss)|32|0.9719|0.1|
|3|112x112|ResNet-18|90.140%|5.986|回归(L2Loss)|32|1.121|0.01|
|4|224x224|ResNet-18|90.064%|8.475|分类(交叉熵)|32|TBD|TBD|
|5|224x224|ResNet-50|90.034%|TBD|分类(交叉熵)|32|TBD|TBD|


### Demo
下载预训练的模型 [Link](https://github.com/foamliu/Age-and-Gender/releases/download/1.0/BEST_checkpoint_.pth.tar)，执行：
```bash
$ python demo.py
```

效果图如下：

原图 | 校准 | 识别 | 标注 |
|---|---|---|---|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/0_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/0_img.jpg)|性别：女, 年龄：29|性别：女, 年龄：24|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/1_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/1_img.jpg)|性别：女, 年龄：29|性别：男, 年龄：26|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/2_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/2_img.jpg)|性别：男, 年龄：34|性别：男, 年龄：49|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/3_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/3_img.jpg)|性别：女, 年龄：29|性别：女, 年龄：29|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/4_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/4_img.jpg)|性别：男, 年龄：36|性别：女, 年龄：23|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/5_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/5_img.jpg)|性别：男, 年龄：29|性别：男, 年龄：15|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/6_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/6_img.jpg)|性别：男, 年龄：34|性别：男, 年龄：32|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/7_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/7_img.jpg)|性别：男, 年龄：42|性别：男, 年龄：42|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/8_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/8_img.jpg)|性别：男, 年龄：36|性别：女, 年龄：31|
|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/9_raw.jpg)|![image](https://github.com/foamliu/Age-and-Gender/raw/master/images/9_img.jpg)|性别：男, 年龄：39|性别：男, 年龄：59|