@page ai_agent ai_agent

# 概述
AI Agent作为AI的代理引擎框架，支持不同推理引擎的注册，对用户提供统一的接口，简化用户操作，目前注册的AI引擎为ucloud_ai云端推理引擎，用户可以根据需要注册其他推理引擎。

## 版权信息
> Apache license v2.0

## 目录结构
```sh
├── include
│   ├── aiagent_common.h         # aiagent模型定义头文件
│   ├── aiagent_engine.h         # aiagent引擎注册结构体定义头文件
│   ├── aiagent_service.h        # aiagent对外服务接口定义头文件
│   ├── engine
│   │   └── ucloud_ai_engine.h   # ucloud ai引擎头文件
│   └── aiconfig.h               # AI引擎配置文件
├── package.yaml                 # 编译配置文件
└── src
    ├── aiagent_engine.c         # aiagent引擎管理代码
    ├── aiagent_service.c        # aiagent服务接口实现
    └── engine
    │   └── ucloud_ai_engine.c   # ucloud ai引擎注册实例
    └── example
        └── aiagent_example.c
```

## 依赖组件
* ai_agent

# 常用配置
```sh
def_config:                      # 组件的可配置项
  CONFIG_UCLOUD_AI_ENGINE_ENABLE: 1
```
> 配置AI引擎为ucloud ai推理引擎

# API说明
@ref aiagent_aos_api

# 使用示例
示例代码参考example/aiagent_example.c，以运行ucloud_ai_demo为例，具体步骤如下：

## 添加示例代码
> ai_agent组件的package.yaml中添加example
```sh
source_file:
  - "example/aiagent_example.c" # add aiagent_example.c
```

> helloworld_demo组件的application_start中添加代码
```sh
    /*init network service*/
    event_service_init(NULL);
    netmgr_service_init(NULL);
```

## 云端功能开通
> 登陆https://vision.aliyun.com免费开通如下功能：
```sh
人脸人体: https://vision.aliyun.com/facebody
文字识别: https://vision.aliyun.com/ocr
分割抠图: https://vision.aliyun.com/imageseg
目标检测: https://vision.aliyun.com/objectdet

如没有阿里云账号，请登陆http://www.aliyun.com开通
```

> 登陆oss.console.aliyun.com创建bucket
```sh
1. 创建Bucket时地域选择“上海”
2. 读写权限选择“公共读”
```
## ucloud_ai/package.yaml中配置OSS信息
```sh
OSS_ACCESS_KEY: "Your-Access-Key"
OSS_ACCESS_SECRET: "Your-Access-Secret"
OSS_ENDPOINT: "Your-OSS-Endpoint"
OSS_BUCKET: "Your-OSS-Bucket"

KEY以及SECRET获取链接: https://usercenter.console.aliyun.com/#/accesskey
ENDPOINT使用默认即可，BUCKET请使用你创建好的Bucket名称
```

## 添加ai_agent组件
> helloworld_demo组件的package.yaml中添加
```sh
depends:
  - ai_agent: dev_aos   # helloworld_demo中引入ai_agent组件
  - netmgr: dev_aos     # helloworld_demo中引入netmgr组件
  - littlefs: dev_aos   # helloworld_demo中引入littlefs组件
```

## 编译
```sh
cd solutions/ucloud_ai_demo && aos make
```

## 烧录固件
> helloworld_demo bin烧录：
```sh
aos burn
```

> littlefs文件系统烧录：
```sh
aos burn -f hardware/chip/haas1000/release/write_flash_tool/ota_bin/littlefs.bin#0xB32000
```
本组件例子中使用到到图片存放在代码中hardware/chip/haas1000/prebuild/data目录，除烧录helloworld demo image外，需烧录littlefs文件系统。

## ucloud_ai示例测试

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 0  # 人脸比对
```

> CLI关键日志：
```sh
Facebody comparing result:
confidence: 84.6
location at x: 159, y: 76, w: 143, h: 194
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 1  # 人物动漫化
```

> CLI关键日志：
```sh
Generate human anime style result:
url: http://vibktprfx-prod-prod-aic-gd-cn-shanghai.oss-cn-shanghai.aliyuncs.com/person-image-cartoonizer/4F10FBB8-19DC-476B-A51D-458CD180A24F_07f2_20210331-142711.jpg?Expires=1617202631&OSSAccessKeyId=LTAI4FoLmvQ9urWXgSRpDvh1&Signature=YCLva1stK27Cn657nMeYhgFt7h4%3D
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 2  # 表情识别
```

> CLI关键日志：
```sh
Recognize expression result:
type: anger, probability: 1.0
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 3  # 目标检测
```

> CLI关键日志：
```sh
Detect object result:
type: potted plant, Score: 0.4, x: 35, y: 274, w: 185, h: 443
Detect object result:
type: potted plant, Score: 0.5, x: 663, y: 121, w: 847, h: 457
Detect object result:
type: monitor, Score: 0.6, x: 250, y: 43, w: 670, h: 439
Detect object result:
type: keyboard, Score: 0.5, x: 239, y: 435, w: 489, h: 485
Detect object result:
type: keyboard, Score: 0.3, x: 241, y: 428, w: 639, h: 506
Detect object result:
type: mouse, Score: 0.2, x: 660, y: 479, w: 750, h: 528
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 4  # 主体检测
```

> CLI关键日志：
```sh
Detect main body result:
main body location x: 1, y: 31, w: 461, h: 528
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 5  # 通用分割
```

> CLI关键日志：
```sh
Segment common image result:
url: http://viapi-cn-shanghai-dha-segmenter.oss-cn-shanghai.aliyuncs.com/upload/result_humansegmenter/2021-3-31/invi_humansegmenter_016172056326493133388_avpTpm.jpg?Expires=1617207432&OSSAccessKeyId=LTAI4FoLmvQ9urWXgSRpDvh1&Signature=WREOAoBUIvpLSyygGxKPfi82HfI%3D
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 6  # 面部分割
```

> CLI关键日志：
```sh
Segment face result:
image url: http://viapi-cn-shanghai-dha-segmenter.oss-cn-shanghai.aliyuncs.com/upload/result_HeadSegmenter/2021-3-31/invi_HeadSegmenter_016172056422754285967296_yRhj6C.png?Expires=1617207442&OSSAccessKeyId=LTAI4FoLmvQ9urWXgSRpDvh1&Signature=wR7GUxqrrwRF9u3mCJ%2BNDhAPCzk%3D
location at x: 153, y: 66, w: 160, h: 209
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 7  # 身份证证明识别
```

> CLI关键日志：
```sh
Recognize identity card face side result:
address: 江苏省南京市浦口区天天小区1栋11号
nationality: 汉
card location: x0: 1126563840, y0: 1126563840
card location: x1: 0, y1: 0
card location: x2: 0, y2: 0
card location: x3: 0, y3: 0
face location: x0: 1126563840, y0: 1126563840
face location: x1: 0, y1: 0
face location: x2: 0, y2: 0
face location: x3: 0, y3: 0
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 8  # 身份证反面识别
```

> CLI关键日志：
```sh
Recognize identity card back side result:
start date: 20131010
issue: 杭州市公安局余杭分局
end date: 20231010
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 9  # 银行卡识别
```

> CLI关键日志：
```sh
bankName: 中国工商银行
cardNumber: 6212262315007683105
validDate: 07/26
Recognize bank card result:
bank name: 中国工商银行
card number: 6212262315007683105
valid date: 07/26
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 10  # 文本识别
```

> CLI关键日志：
```sh
Recognize character result:
text: 《爆炸新闻》
probability: 1.0
text area: left: 141, top: 442, weight: 89, height: 20
Recognize character result:
text: 19
probability: 1.0
text area: left: 115, top: 173, weight: 100, height: 109
Recognize character result:
text: 豆瓣评分7.1
probability: 0.9
text area: left: 168, top: 447, weight: 98, height: 18
Recognize character result:
text: 2019||美国|加拿大|剧情|传记
probability: 0.5
text area: left: 181, top: 450, weight: 241, height: 17
Recognize character result:
text: 杰伊·罗奇 导演
probability: 0.9
text area: left: 216, top: 454, weight: 112, height: 18
Recognize character result:
text: 2021年2月星期五
probability: 1.0
text area: left: 231, top: 187, weight: 150, height: 18
Recognize character result:
text: 农历正月初八
probability: 1.0
text area: left: 261, top: 193, weight: 104, height: 18
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 11  # 垃圾分类
```

> CLI关键日志：
```sh
Recognize rubbish result:
rubbish: 塑料饮料瓶
rubbish score: 1.0
category: 可回收垃圾
category score: 1.0
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 12  # 水果检测
```

> CLI关键日志：
```sh
Recognize fruits result:
fruit name: 橙子
fruit score: 0.8
fruit location: x: 214, y: 44, w: 509, h: 438
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 13  # 图像人体擦除
```

> CLI关键日志：
```sh
Erase person result:
url: http://algo-app-isr-lab-cn-shanghai-prod.oss-cn-shanghai.aliyuncs.com/remove-person/2021-03-31_15%3A34%3A17.959810_person_org.jpg?Expires=1617206658&OSSAccessKeyId=LTAI4FoLmvQ9urWXgSRpDvh1&Signature=jFwwnSYOw2fXev1%2Fm5az4JBa%2B5Q%3D
aiagent comp test success!
```

### CLI命令行输入：
```sh
aiagent -e ucloud-ai -m 14  # 风格迁移
```

> CLI关键日志：
```sh
Extend image style result:
out image url: http://viapi-cn-shanghai-dha-filter.oss-cn-shanghai.aliyuncs.com/upload/result_/2021-3-31/invi__016172056608421021961_1mdK2W.jpg?Expires=1617207460&OSSAccessKeyId=LTAI4FoLmvQ9urWXgSRpDvh1&Signature=3t8Q%2BwTAo6Lib9utaoMrWUJX6Iw%3D
aiagent comp test success!
```

# 注意事项
NA

# FAQ
NA

如果遇到：
> error code: InvalidAction.NotFound
请确认Endpoint是否设置正确

> error code: InvalidApi.NotPurchase

请确认是否在vision.aliyun.com开通相应功能

> tlefs_vfs.c:677:error: lfs_vfs_open /data/ucloud_ai_image/object.jpg failed, ret - -2

请确认是否有拷贝图片到hardware/chip/haas1000/prebuild/data/ucloud_ai_image目录。
如果有，可能是文件系统存储空间不够，此时请执行：
1. 在ucloud_ai/package.yaml中屏蔽prebuild_script: cp_resources.py；
2. 删除ucloud_ai_image中其他文件，仅保留测试case使用的图片文件；
3. 再重新编译烧录。