開發與維運

人臉識別實戰教程

2020-06-02

人臉識別實戰教程

本教程介紹如何使用Alibaba Cloud SDK for Java進行人體屬性識別、人體計數和人臉搜索任務。

背景信息

人臉人體識別技術是基於阿里雲深度學習算法，結合圖像或視頻的人臉檢測、分析、比對以及人體檢測等技術，為您提供人臉人體的檢測定位、人臉屬性識別和人臉比對等能力。可以為開發者和企業提供高性能的在線API服務，應用於人臉AR、人臉識別和認證、大規模人臉檢索、照片管理等各種場景。

前提條件

在開始之前，請確保完成以下步驟：

  1. 開通人臉人體能力，請參見上述開發前準備。
    image.png
  2. 在您的Java工程中添加人臉人體能力的pom依賴：
<!-- https://mvnrepository.com/artifact/com.aliyun/aliyun-java-sdk-facebody -->
<dependency>
    <groupId>com.aliyun</groupId>
    <artifactId>aliyun-java-sdk-facebody</artifactId>
    <version>1.0.8</version>
</dependency>

人臉屬性識別

RecognizeFace可以識別檢測人臉的性別、年齡、表情、眼鏡四種屬性。返回人臉的1024維深度學習特徵，基於這個特徵並按照特徵比較規則，您可實現高性能的人臉識別。
例如需要識別以下圖片中的人臉屬性。
image.png
示例代碼如下：

import com.aliyun.CommonConfig;
import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.facebody.model.v20191230.RecognizeFaceRequest;
import com.aliyuncs.facebody.model.v20191230.RecognizeFaceResponse;
import com.aliyuncs.profile.DefaultProfile;
import com.google.gson.Gson;

public class RecognizeFaceDemo {
    private static DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", "<access key id>", "<access key secret>");
    private static IAcsClient client = new DefaultAcsClient(profile);

    public static void main(String[] args) {
        String recognizeFaceURL = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/recognize_1.jpg";
        recognizeFace(recognizeFaceURL);
    }
    /**
     * 人臉屬性識別
     * @param imageURL 圖片URL地址
     */
    private static void recognizeFace(String imageURL)
    {
        RecognizeFaceRequest recognizeFaceRequest = new RecognizeFaceRequest();
        recognizeFaceRequest.setImageURL(imageURL);

        try {
            RecognizeFaceResponse recognizeFaceResponse = client.getAcsResponse(recognizeFaceRequest);
            System.out.println("人臉屬性識別：");
            System.out.println(new Gson().toJson(recognizeFaceResponse));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

代碼返回結果類似如下：

{
    "requestId": "9E461AFC-B125-4B04-8644-4D39C495B61A",
    "data": {
        "faceCount": 1,
        "landmarkCount": 105,
        "denseFeatureLength": 1024,
        "faceRectangles": [
            604,
            124,
            128,
            160
        ],
        "faceProbabilityList": [
            55.0
        ],
        "poseList": [
            -8.399615,
            -10.90538,
            14.451606
        ],
        "landmarks": [
            614.1295,
            161.77643,
            647.84894,
            ...
        ],
        "pupils": [
            637.25793,
            177.63129,
            5.2395487,
            689.71075,
            192.1905,
            5.2395487
        ],
        "genderList": [
            0
        ],
        "ageList": [
            24
        ],
        "expressions": [
            1
        ],
        "glasses": [
            1
        ],
        "denseFeatures": [
            "-0.0080740926787257195",
            "0.011421392671763897",
            "-0.027754634618759155",
            ...
        ]
    }
}

從返回結果中得到的該圖片人臉屬性識別結果如下：

  • 人臉個數：1
  • 性別：女性
  • 年齡：24
  • 是否微笑：微笑
  • 是否戴眼鏡：有眼鏡

人體計數

DetectBodyCount可以檢測輸入圖片中人體的個數。
待檢測圖片如下：
image.png
示例代碼如下：

import com.aliyun.CommonConfig;
import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.facebody.model.v20191230.DetectBodyCountRequest;
import com.aliyuncs.facebody.model.v20191230.DetectBodyCountResponse;
import com.aliyuncs.profile.DefaultProfile;
import com.google.gson.Gson;

/**
 * 人體計數
 */
public class DetectBodyCountDemo {
    private static DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", "<access key id>", "<access key secret>");
    private static IAcsClient client = new DefaultAcsClient(profile);

    public static void main(String[] args) {
        String detectBodyCountURL = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/multiplayer.jpg";
        detectBodyCount(detectBodyCountURL);
    }

    /**
     * 人體計數
     * @param imageURL 圖片URL地址
     */
    private static void detectBodyCount(String imageURL)
    {
        DetectBodyCountRequest detectBodyCountRequest = new DetectBodyCountRequest();
        detectBodyCountRequest.setImageURL(imageURL);

        try {
            DetectBodyCountResponse detectBodyCountResponse = client.getAcsResponse(detectBodyCountRequest);
            System.out.println("人體計數：");
            System.out.println(new Gson().toJson(detectBodyCountResponse));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

代碼返回結果類似如下：

{
    "requestId": "15170328-DF69-4F24-900D-9915645A4237",
    "data": {
        "personNumber": 3
    }
}

從返回結果中得到該圖片被檢測到有3個人。

人臉搜索

人臉搜索可以根據輸入圖片，在數據庫中搜索相似的人臉圖片數據。
人臉搜索操作流程示意圖如下：
image.png
說明: 公測階段每個阿里雲主賬號只能創建一個數據庫，且開通服務後默認創建一個名稱為default的數據庫，每個數據庫最多可以添加5萬張圖片。
涉及的API如下：

例如我們有以下人臉數據：
image.png
將上面圖片樣本加入人臉庫，使用下面人臉數據搜索人臉庫。
image.png
示例代碼如下：

import com.aliyun.CommonConfig;
import com.aliyuncs.DefaultAcsClient;
import com.aliyuncs.IAcsClient;
import com.aliyuncs.exceptions.ClientException;
import com.aliyuncs.exceptions.ServerException;
import com.aliyuncs.facebody.model.v20191230.*;
import com.aliyuncs.profile.DefaultProfile;
import com.google.gson.Gson;

public class SearchFaceDemo {
    private static DefaultProfile profile = DefaultProfile.getProfile("cn-shanghai", CommonConfig.ACCESSKEY_ID, CommonConfig.ACCESSKEY_SECRET);
    private static IAcsClient client = new DefaultAcsClient(profile);

    public static void main(String[] args) throws InterruptedException {

        String dbName = "default";

        String human1_1 = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/human_11.jpg";
        String human1_2 = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/human_12.jpg";
        String human2_1 = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/human_21.jpg";
        String human2_2 = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/human_22.jpg";
        String sample = "https://visionapi-test.oss-cn-shanghai.aliyuncs.com/sample.jpg";

        String entityId1 = "human1";
        String entityId2 = "human2";

        // 1.創建人臉樣本
        addFaceEntity(dbName, entityId1);
        addFaceEntity(dbName, entityId2);

        // 2.向人臉樣本中加入人臉，每個樣本人臉上限為5
        addFace(dbName,entityId1,human1_1);
        addFace(dbName,entityId1,human1_2);
        addFace(dbName,entityId2,human2_1);
        addFace(dbName,entityId2,human2_2);

        Thread.currentThread().sleep(3000);

        // 3.到人臉庫中查找
        searchFace(dbName,sample,1);
    }

    /**
     * 添加人臉樣本
     * @param dbName 數據庫名稱
     * @param entityId 實體ID
     */
    private static void addFaceEntity(String dbName, String entityId)
    {
        AddFaceEntityRequest addFaceEntityRequest = new AddFaceEntityRequest();
        addFaceEntityRequest.setDbName(dbName);
        addFaceEntityRequest.setEntityId(entityId);
        try{
            AddFaceEntityResponse addFaceEntityResponse = client.getAcsResponse(addFaceEntityRequest);
            System.out.println("添加人臉樣本：");
            System.out.println(new Gson().toJson(addFaceEntityResponse));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }

    /**
     * 添加人臉數據
     * @param dbName 數據庫名稱
     * @param entityId 實體ID
     * @param imageUrl 人臉圖片地址，必須是同Region的OSS的圖片地址。人臉必須是正面無遮擋單人人臉。
     */
    private static void addFace(String dbName, String entityId, String imageUrl)
    {
        AddFaceRequest addFaceRequest = new AddFaceRequest();
        addFaceRequest.setDbName(dbName);
        addFaceRequest.setEntityId(entityId);
        addFaceRequest.setImageUrl(imageUrl);

        try{
            AddFaceResponse addFaceResponse = client.getAcsResponse(addFaceRequest);
            System.out.println("添加人臉數據：");
            System.out.println(new Gson().toJson(addFaceResponse));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }

    /**
     * 搜索人臉
     * @param dbName 數據庫名稱
     * @param imageUrl 圖片URL地址。必須是同Region的OSS地址。
     */
    private static void searchFace(String dbName, String imageUrl, Integer limit)
    {
        SearchFaceRequest searchFaceRequest = new SearchFaceRequest();
        searchFaceRequest.setDbName(dbName);
        searchFaceRequest.setImageUrl(imageUrl);
        searchFaceRequest.setLimit(limit);
        try{
            SearchFaceResponse searchFaceResponse = client.getAcsResponse(searchFaceRequest);
            System.out.println("搜索人臉：");
            System.out.println(new Gson().toJson(searchFaceResponse));
        } catch (ServerException e) {
            e.printStackTrace();
        } catch (ClientException e) {
            System.out.println("ErrCode:" + e.getErrCode());
            System.out.println("ErrMsg:" + e.getErrMsg());
            System.out.println("RequestId:" + e.getRequestId());
        }
    }
}

返回結果類似如下：

添加人臉樣本：
{"requestId":"718AE1B9-F1B3-4720-98FA-69E2A41F4C31"}
添加人臉樣本：
{"requestId":"9CE55AD1-1E21-49F0-BB2C-B3C07A015EAE"}
添加人臉數據：
{"requestId":"9551C7C7-8027-4E82-82FB-15FEC0EB8161","data":{"faceId":"1589791311965000"}}
添加人臉數據：
{"requestId":"C193A2E0-F22A-4418-AAD3-F28749BDDC17","data":{"faceId":"1589791312203000"}}
添加人臉數據：
{"requestId":"687D9D8B-5EDE-4774-B377-C1CB4E0FB855","data":{"faceId":"1589791312433000"}}
添加人臉數據：
{"requestId":"1E0948E3-7D75-4709-A624-073B4CB54874","data":{"faceId":"1589791312679000"}}
搜索人臉：
{
    "requestId": "F0543F2A-216F-495A-8B4D-20CA06B605A7",
    "data": {
        "matchList": [
            {
                "faceItems": [
                    {
                        "faceId": "1589791312203000",
                        "score": 0.86908734,
                        "extraData": "",
                        "entityId": "human1"
                    }
                ],
                "location": {
                    "x": 592,
                    "y": 132,
                    "width": 152,
                    "height": 204
                }
            }
        ]
    }
}

從返回結果中得出輸入圖片：

  • 被匹配到的樣本數據faceId為1589791312203000。
  • 匹配到的樣本名稱為human1。
  • 與樣本數據相似度為0.86908734。

Leave a Reply

Your email address will not be published. Required fields are marked *