你当前正在访问 Microsoft Azure Global Edition 技术文档网站。 如果需要访问由世纪互联运营的 Microsoft Azure 中国技术文档网站,请访问 https://docs.azure.cn

Large Person Group Person - Add Face From Stream

将某个人员的人脸添加到大型人员组以进行人脸识别或验证。 若要处理包含多个人脸的图像,可以将输入人脸指定为具有 targetFace 矩形的图像。 它返回一个 persistedFaceId,表示添加的人脸。 不会存储任何图像。 只有提取的人脸特征将存储在服务器上,直到调用 LargePersonGroup PersonFace - DeleteLargePersonGroup Person - DeleteLargePersonGroup - Delete
注意 persistedFaceId 不同于 Face - Detect 生成的 faceId

  • 人脸图像质量越高,识别精度就越高。 请考虑高质量的人脸:正面、清晰,人脸大小为 200x200 像素 (100 像素,眼睛之间) 或更大。
  • 每人最多可以容纳 248 张人脸。
  • 支持 JPEG、PNG、GIF (第一帧) 和 BMP 格式。 允许的图像文件大小为 1KB 到 6MB。
  • “targetFace”矩形应包含一张人脸。 零张或多张人脸将被视为错误。 如果提供的“targetFace”矩形未从 人脸 - 检测返回,则不能保证成功检测和添加人脸。
  • (36x36 - 4096x4096 像素) 、大头姿势或大遮挡会导致故障。
  • 向/从同一个人添加/删除人脸将按顺序处理。 向/从不同人员添加/删除人脸是并行处理的。
  • 图像中最小可检测人脸大小为 36x36 像素,且不超过 1920x1080 像素。 尺寸高于 1920x1080 像素的图像需要成比例更大的最小人脸大小。
  • 可以提供不同的“detectionModel”值。 若要使用和比较不同的检测模型,请参阅 如何指定检测模型
POST {Endpoint}/face/v1.0/largepersongroups/{largePersonGroupId}/persons/{personId}/persistedfaces
POST {Endpoint}/face/v1.0/largepersongroups/{largePersonGroupId}/persons/{personId}/persistedfaces?overload=stream&userData={userData}&targetFace={targetFace}&detectionModel={detectionModel}

URI 参数

名称 必需 类型 说明
Endpoint
path True

string

支持的认知服务终结点 (协议和主机名,例如: https://westus.api.cognitive.microsoft.com) 。

largePersonGroupId
path True

string

引用特定大型人员组的 ID。

Regex pattern: ^[a-z0-9-_]+$

personId
path True

string

uuid

引用特定人员的 ID。

detectionModel
query

DetectionModel

检测模型的名称。 检测模型用于检测提交图像中的人脸。 执行人脸 - 检测或 (大型) 人脸列表 - 添加人脸或 (大型) PersonGroup - 添加人脸时,可以提供检测模型名称。 默认值为“detection_01”,如果需要其他模型,请显式指定它。

targetFace
query

integer[]

一个人脸矩形,用于指定要以“targetFace=left,top,width,height”格式添加到人员的目标人脸。 例如“targetFace=10,10,100,100”。 如果图像中有多个人脸,则需要 targetFace 来指定要添加的人脸。 没有 targetFace 意味着在整个图像中只检测到一张人脸。

userData
query

string

用户指定用于任何目的的人脸数据。 最大长度为 1KB。

请求头

Media Types: "application/octet-stream"

名称 必需 类型 说明
Ocp-Apim-Subscription-Key True

string

请求正文

Media Types: "application/octet-stream"

名称 类型 说明
Image

object

图像流。

响应

名称 类型 说明
200 OK

PersistedFace

成功调用将返回新的 persistedFaceId。

Other Status Codes

APIError

错误响应。

安全性

Ocp-Apim-Subscription-Key

Type: apiKey
In: header

示例

Add face to person with stream example

Sample Request

POST {Endpoint}/face/v1.0/largepersongroups/abc/persons/4caa25ee-3bc6-4e88-adf8-12455ce7aab0/persistedfaces?overload=stream&userData={customized User data}&targetFace=10,10,100,100&detectionModel=detection_01


"{Image stream in base 64 encoded format}"

Sample Response

{
  "persistedFaceId": "6e04c175-219e-42a2-9d26-0e7b790e1ef4"
}

定义

名称 说明
APIError

API 返回的错误信息

DetectionModel

检测模型的名称。 检测模型用于检测提交图像中的人脸。 执行人脸 - 检测或 (大型) 人脸列表 - 添加人脸或 (大型) PersonGroup - 添加人脸时,可以提供检测模型名称。 默认值为“detection_01”,如果需要其他模型,请显式指定它。

Error

错误正文。

PersistedFace

PersonFace 对象。

APIError

API 返回的错误信息

名称 类型 说明
error

Error

错误正文。

DetectionModel

检测模型的名称。 检测模型用于检测提交图像中的人脸。 执行人脸 - 检测或 (大型) 人脸列表 - 添加人脸或 (大型) PersonGroup - 添加人脸时,可以提供检测模型名称。 默认值为“detection_01”,如果需要其他模型,请显式指定它。

名称 类型 说明
detection_01

string

detection_02

string

detection_03

string

Error

错误正文。

名称 类型 说明
code

string

message

string

PersistedFace

PersonFace 对象。

名称 类型 说明
persistedFaceId

string

目标人脸的 persistedFaceId,该持久化且不会过期。 不同于 Face - Detect 创建的 faceId,将在检测调用后 faceIdTimeToLive 指定的时间过期。

userData

string

用户提供的数据附加到人脸。 大小限制为 1KB。