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

IFaceOperations 接口

定义

FaceOperations 操作。

public interface IFaceOperations
type IFaceOperations = interface
Public Interface IFaceOperations
派生

方法

DetectWithStreamWithHttpMessagesAsync(Stream, Nullable<Boolean>, Nullable<Boolean>, IList<FaceAttributeType>, String, Nullable<Boolean>, String, Nullable<Int32>, Dictionary<String,List<String>>, CancellationToken)

FaceOperations 操作。

DetectWithUrlWithHttpMessagesAsync(String, Nullable<Boolean>, Nullable<Boolean>, IList<FaceAttributeType>, String, Nullable<Boolean>, String, Nullable<Int32>, Dictionary<String,List<String>>, CancellationToken)

FaceOperations 操作。

FindSimilarWithHttpMessagesAsync(Guid, String, String, IList<Nullable<Guid>>, Nullable<Int32>, FindSimilarMatchMode, Dictionary<String,List<String>>, CancellationToken)

给定查询人脸的 faceId,用于在 faceId 数组、人脸列表或大型人脸列表中搜索类似的人脸。 faceId 数组包含人 脸 - 检测创建的人脸,这将在创建后 24 小时过期。 “faceListId”由 FaceList - 创建 包含不会过期的 persistedFaceIds 创建。 “largeFaceListId”由 LargeFaceList - 创建 ,其中包含也不会过期的 persistedFaceId。 根据输入,返回的相似人脸列表包含按相似性排名的 faceId 或 persistedFaceId。 <br/>Find similar 有两种工作模式:“matchPerson”和“matchFace”。 “matchPerson”是默认模式,它尝试通过使用内部同一人员阈值来查找同一个人的人脸。 查找已知人员的其他照片很有用。 请注意,如果没有人脸通过内部阈值,将返回空列表。 “matchFace”模式忽略同一人阈值,并返回排名相似的人脸,即使相似性较低。 它可用于搜索名人面孔等情况。 <br/>与查询人脸的 faceId 关联的“recognitionModel”应与目标 faceId 数组、人脸列表或大型人脸列表使用的“recognitionModel”相同。

GroupWithHttpMessagesAsync(IList<Guid>, Dictionary<String,List<String>>, CancellationToken)

根据人脸相似性将候选人脸划分为多个组。<Br/>

  • 输出是一个或多个不连续的人脸组和一个乱组。 人脸组包含外观相似的人脸,通常是同一个人。 人脸组按组大小(即人脸数量)进行排名。 请注意,属于同一人的人脸可能会在结果中拆分为多个组。
  • MessyGroup 是一个特殊的人脸组,其中包含无法从原始人脸中找到任何相似的对应人脸的人脸。 如果所有人脸都找到了对应人,则结果中不会显示 messyGroup。
  • 组 API 至少需要 2 张候选人脸,最多需要 1000 张。 建议尝试 人脸 - 验证 是否只有 2 张候选人脸。
  • 与查询人脸关联的“recognitionModel”faceIds 应相同。
IdentifyWithHttpMessagesAsync(IList<Guid>, String, String, String, IList<String>, Nullable<Int32>, Nullable<Double>, Dictionary<String, List<String>>, CancellationToken)

FaceOperations 操作。

VerifyFaceToFaceWithHttpMessagesAsync(Guid, Guid, Dictionary<String,List<String>>, CancellationToken)

验证两张人脸是否属于同一个人,或者一张人脸是否属于某一个人。 <br/> 备注:<br />

  • 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) 或更大。
  • 对于对准确性敏感的方案,请自行做出判断。
  • 与查询人脸关联的“recognitionModel”faceIds 应与目标人脸、人员组或大型人员组使用的“recognitionModel”相同。
VerifyFaceToPersonWithHttpMessagesAsync(Guid, Guid, String, String, Dictionary<String,List<String>>, CancellationToken)

验证两张人脸是否属于同一个人。 将人脸 ID 与人员 ID 进行比较

扩展方法

DetectWithStreamAsync(IFaceOperations, Stream, Nullable<Boolean>, Nullable<Boolean>, IList<FaceAttributeType>, String, Nullable<Boolean>, String, Nullable<Int32>, CancellationToken)

FaceOperations 操作。

DetectWithUrlAsync(IFaceOperations, String, Nullable<Boolean>, Nullable<Boolean>, IList<FaceAttributeType>, String, Nullable<Boolean>, String, Nullable<Int32>, CancellationToken)

FaceOperations 操作。

FindSimilarAsync(IFaceOperations, Guid, String, String, IList<Nullable<Guid>>, Nullable<Int32>, FindSimilarMatchMode, CancellationToken)

给定查询人脸的 faceId,用于在 faceId 数组、人脸列表或大型人脸列表中搜索类似的人脸。 faceId 数组包含人 脸 - 检测创建的人脸,这将在创建后 24 小时过期。 “faceListId”由 FaceList - 创建 包含不会过期的 persistedFaceIds 创建。 “largeFaceListId”由 LargeFaceList - 创建 ,其中包含也不会过期的 persistedFaceId。 根据输入,返回的相似人脸列表包含按相似性排名的 faceId 或 persistedFaceId。 <br/>Find similar 有两种工作模式:“matchPerson”和“matchFace”。 “matchPerson”是默认模式,它尝试通过使用内部同一人员阈值来查找同一个人的人脸。 查找已知人员的其他照片很有用。 请注意,如果没有人脸通过内部阈值,将返回空列表。 “matchFace”模式忽略同一人阈值,并返回排名相似的人脸,即使相似性较低。 它可用于搜索名人面孔等情况。 <br/>与查询人脸的 faceId 关联的“recognitionModel”应与目标 faceId 数组、人脸列表或大型人脸列表使用的“recognitionModel”相同。

GroupAsync(IFaceOperations, IList<Guid>, CancellationToken)

根据人脸相似性将候选人脸划分为多个组。<Br/>

  • 输出是一个或多个不连续的人脸组和一个乱组。 人脸组包含外观相似的人脸,通常是同一个人。 人脸组按组大小(即人脸数量)进行排名。 请注意,属于同一人的人脸可能会在结果中拆分为多个组。
  • MessyGroup 是一个特殊的人脸组,其中包含无法从原始人脸中找到任何相似的对应人脸的人脸。 如果所有人脸都找到了对应人,则结果中不会显示 messyGroup。
  • 组 API 至少需要 2 张候选人脸,最多需要 1000 张。 建议尝试 人脸 - 验证 是否只有 2 张候选人脸。
  • 与查询人脸关联的“recognitionModel”faceIds 应相同。
IdentifyAsync(IFaceOperations, IList<Guid>, String, String, String, IList<String>, Nullable<Int32>, Nullable<Double>, CancellationToken)

FaceOperations 操作。

VerifyFaceToFaceAsync(IFaceOperations, Guid, Guid, CancellationToken)

验证两张人脸是否属于同一个人,或者一张人脸是否属于某一个人。 <br/> 备注:<br />

  • 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) 或更大。
  • 对于对准确性敏感的方案,请自行做出判断。
  • 与查询人脸关联的“recognitionModel”faceIds 应与目标人脸、人员组或大型人员组使用的“recognitionModel”相同。
VerifyFaceToPersonAsync(IFaceOperations, Guid, Guid, String, String, CancellationToken)

验证两张人脸是否属于同一个人。 将人脸 ID 与人员 ID 进行比较

适用于