臉部偵測教學課程

目標

在本範例中，您將使用 Google Vision API 偵測圖片中的臉孔。為證明系統正確偵測到臉部，您接著會使用該資料在每個臉部周圍繪製方塊。

查看程式碼範例：本教學課程的程式碼範例大多擷取自 GitHub 中的大型程式碼檔案。只要按一下範例上方的「前往 GitHub 查看」按鈕，即可查看及下載程式碼範例原屬的完整檔案。

費用

在本文件中，您會使用下列 Google Cloud的計費元件：

• Cloud Vision

如要根據預測用量估算費用，請使用 Pricing Calculator。

初次使用 Google Cloud 的使用者可能符合免費試用期資格。

事前準備

登入 Google Cloud 帳戶。如果您是 Google Cloud新手，歡迎 建立帳戶，親自評估產品在實際工作環境中的成效。新客戶還能獲得價值 $300 美元的免費抵免額，可用於執行、測試及部署工作負載。

In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

Roles required to select or create a project

• Select a project: Selecting a project doesn't require a specific IAM role—you can select any project that you've been granted a role on.
• Create a project: To create a project, you need the Project Creator role (roles/resourcemanager.projectCreator), which contains the resourcemanager.projects.create permission. Learn how to grant roles.

Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

Go to project selector

Verify that billing is enabled for your Google Cloud project.

Enable the Google Cloud Vision API.

Roles required to enable APIs

To enable APIs, you need the Service Usage Admin IAM role (roles/serviceusage.serviceUsageAdmin), which contains the serviceusage.services.enable permission. Learn how to grant roles.

Enable the API

In the Google Cloud console, on the project selector page, select or create a Google Cloud project.

Roles required to select or create a project

• Select a project: Selecting a project doesn't require a specific IAM role—you can select any project that you've been granted a role on.
• Create a project: To create a project, you need the Project Creator role (roles/resourcemanager.projectCreator), which contains the resourcemanager.projects.create permission. Learn how to grant roles.

Note: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.

Go to project selector

Verify that billing is enabled for your Google Cloud project.

Enable the Google Cloud Vision API.

Roles required to enable APIs

To enable APIs, you need the Service Usage Admin IAM role (roles/serviceusage.serviceUsageAdmin), which contains the serviceusage.services.enable permission. Learn how to grant roles.

Enable the API

1. 設定環境，以便使用 應用程式預設憑證。
2. 設定語言專屬工作和工具：

   C#

   • 安裝 Google 用戶端程式庫
   • 安裝 Visual Studio 2012/2015。

   Java

   • 安裝 Java。
   • API 參考資料。

   • 下載並安裝 Apache Maven 建構系統。Maven 會確保在建構專案時安裝 Google API 用戶端程式庫和 Vision API 用戶端程式庫，因為我們已將這些程式庫納入 pom.xml。

     <dependency>
       <groupId>com.google.apis</groupId>
       <artifactId>google-api-services-vision</artifactId>
       <version>v1-rev20231219-2.0.0</version>
     </dependency>
     <dependency>
       <groupId>com.google.auth</groupId>
       <artifactId>google-auth-library-oauth2-http</artifactId>
     </dependency>
     <dependency>
       <groupId>com.google.guava</groupId>
       <artifactId>guava</artifactId>
     </dependency>
     <dependency>
       <groupId>com.google.http-client</groupId>
       <artifactId>google-http-client-jackson2</artifactId>
     </dependency>

     ...

   Node.js

   • 安裝 Google 用戶端程式庫
   • 安裝 node.js。
   • API 參考資料。

   • 安裝 npm 和 node-canvas。程式碼範例包含 package.json，可使用 npm install 指令安裝所有依附元件。請注意，node-canvas 有其他可能需要安裝的依附元件，詳情請參閱 node-canvas 安裝文件。

     {
       "name": "nodejs-docs-samples-vision",
       "private": true,
       "license": "Apache-2.0",
       "author": "Google LLC",
       "engines": {
         "node": ">=16.0.0"
       },
       "files": [
         "*.js"
       ],
       "scripts": {
         "test": "c8 mocha -p -j 2 system-test --timeout 600000"
       },
       "dependencies": {
         "@google-cloud/vision": "^4.0.0",
         "natural": "^8.0.0",
         "pureimage": "^0.3.17",
         "redis": "^4.6.5",
         "yargs": "^17.7.1"
       },
       "devDependencies": {
         "@google-cloud/storage": "^7.0.0",
         "@types/uuid": "^10.0.0",
         "@types/yargs": "^17.0.22",
         "c8": "^10.0.0",
         "chai": "^4.5.0",
         "mocha": "^10.2.0",
         "uuid": "^10.0.0"
       }
     }
     

   PHP

   • 安裝 Google 用戶端程式庫
   • 安裝 PHP 和 Composer。
   • API 參考資料。
   • 安裝並啟用 PHP 適用的 GD。

   Python

   • 安裝 Google 用戶端程式庫
   • 安裝 Python。
   • API 參考資料。
   • 安裝 pillow。

   Ruby

   • 安裝 Google 用戶端程式庫
   • 安裝 Ruby 和 Bundler。
   • API 參考資料。
   • 安裝 ImageMagick。

建立服務物件

如要使用官方用戶端 SDK 存取 Google API，請根據 API 的導覽文件建立服務物件，向 SDK 說明 API。您必須使用憑證，從 Vision API 的探索服務擷取該檔案：

Java

import com.google.api.client.googleapis.javanet.GoogleNetHttpTransport;
import com.google.api.client.json.JsonFactory;
import com.google.api.client.json.gson.GsonFactory;
import com.google.api.services.vision.v1.Vision;
import com.google.api.services.vision.v1.VisionScopes;
import com.google.api.services.vision.v1.model.AnnotateImageRequest;
import com.google.api.services.vision.v1.model.AnnotateImageResponse;
import com.google.api.services.vision.v1.model.BatchAnnotateImagesRequest;
import com.google.api.services.vision.v1.model.BatchAnnotateImagesResponse;
import com.google.api.services.vision.v1.model.FaceAnnotation;
import com.google.api.services.vision.v1.model.Feature;
import com.google.api.services.vision.v1.model.Image;
import com.google.api.services.vision.v1.model.Vertex;
import com.google.auth.http.HttpCredentialsAdapter;
import com.google.auth.oauth2.GoogleCredentials;
import com.google.common.collect.ImmutableList;
import java.awt.BasicStroke;
import java.awt.Color;
import java.awt.Graphics2D;
import java.awt.Polygon;
import java.awt.image.BufferedImage;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.security.GeneralSecurityException;
import java.util.List;
import javax.imageio.ImageIO;

/** Connects to the Vision API using Application Default Credentials. */
public static Vision getVisionService() throws IOException, GeneralSecurityException {
  GoogleCredentials credential =
      GoogleCredentials.getApplicationDefault().createScoped(VisionScopes.all());
  JsonFactory jsonFactory = GsonFactory.getDefaultInstance();
  return new Vision.Builder(
          GoogleNetHttpTransport.newTrustedTransport(),
          jsonFactory,
          new HttpCredentialsAdapter(credential))
      .setApplicationName(APPLICATION_NAME)
      .build();
}

Node.js

// By default, the client will authenticate using the service account file
// specified by the GOOGLE_APPLICATION_CREDENTIALS environment variable and use
// the project specified by the GCLOUD_PROJECT environment variable. See
// https://googlecloudplatform.github.io/gcloud-node/#/docs/google-cloud/latest/guides/authentication
const vision = require('@google-cloud/vision');

// Creates a client
const client = new vision.ImageAnnotatorClient();

const fs = require('fs');

Python

from google.cloud import vision
from PIL import Image, ImageDraw

client = vision.ImageAnnotatorClient()

傳送臉部偵測要求

如要建構 Vision API 要求，請先參閱 API 說明文件。在這種情況下，您會要求 images 資源annotate您的圖片。對這個 API 提出的要求會以物件形式呈現，並包含 requests 清單。這份清單中的每個項目都包含兩項資訊：

• Base64 編碼的圖片資料
• 您希望系統為該圖片註解的功能清單。

在本例中，您只需對一張圖片要求 FACE_DETECTION 註解，並傳回回應的相關部分：

Java

/** Gets up to {@code maxResults} faces for an image stored at {@code path}. */
public List<FaceAnnotation> detectFaces(Path path, int maxResults) throws IOException {
  byte[] data = Files.readAllBytes(path);

  AnnotateImageRequest request =
      new AnnotateImageRequest()
          .setImage(new Image().encodeContent(data))
          .setFeatures(
              ImmutableList.of(
                  new Feature().setType("FACE_DETECTION").setMaxResults(maxResults)));
  Vision.Images.Annotate annotate =
      vision
          .images()
          .annotate(new BatchAnnotateImagesRequest().setRequests(ImmutableList.of(request)));
  // Due to a bug: requests to Vision API containing large images fail when GZipped.
  annotate.setDisableGZipContent(true);

  BatchAnnotateImagesResponse batchResponse = annotate.execute();
  assert batchResponse.getResponses().size() == 1;
  AnnotateImageResponse response = batchResponse.getResponses().get(0);
  if (response.getFaceAnnotations() == null) {
    throw new IOException(
        response.getError() != null
            ? response.getError().getMessage()
            : "Unknown error getting image annotations");
  }
  return response.getFaceAnnotations();
}

Node.js

async function detectFaces(inputFile) {
  // Make a call to the Vision API to detect the faces
  const request = {image: {source: {filename: inputFile}}};
  const results = await client.faceDetection(request);
  const faces = results[0].faceAnnotations;
  const numFaces = faces.length;
  console.log(`Found ${numFaces} face${numFaces === 1 ? '' : 's'}.`);
  return faces;
}

Python

def detect_face(face_file, max_results=4):
    """Uses the Vision API to detect faces in the given file.

    Args:
        face_file: A file-like object containing an image with faces.

    Returns:
        An array of Face objects with information about the picture.
    """
    client = vision.ImageAnnotatorClient()

    content = face_file.read()
    image = vision.Image(content=content)

    return client.face_detection(image=image, max_results=max_results).face_annotations

處理回應

恭喜！您已偵測到圖片中的臉孔。臉部註解要求的回應包含偵測到的臉部中繼資料，包括涵蓋臉部的多邊形座標。不過，此時這只是一串數字。讓我們使用這些標記，確認您確實已在圖片中找到臉孔。我們將使用 Vision API 傳回的座標，在圖片副本上繪製多邊形：

Java

/** Reads image {@code inputPath} and writes {@code outputPath} with {@code faces} outlined. */
private static void writeWithFaces(Path inputPath, Path outputPath, List<FaceAnnotation> faces)
    throws IOException {
  BufferedImage img = ImageIO.read(inputPath.toFile());
  annotateWithFaces(img, faces);
  ImageIO.write(img, "jpg", outputPath.toFile());
}

/** Annotates an image {@code img} with a polygon around each face in {@code faces}. */
public static void annotateWithFaces(BufferedImage img, List<FaceAnnotation> faces) {
  for (FaceAnnotation face : faces) {
    annotateWithFace(img, face);
  }
}

/** Annotates an image {@code img} with a polygon defined by {@code face}. */
private static void annotateWithFace(BufferedImage img, FaceAnnotation face) {
  Graphics2D gfx = img.createGraphics();
  Polygon poly = new Polygon();
  for (Vertex vertex : face.getFdBoundingPoly().getVertices()) {
    poly.addPoint(vertex.getX(), vertex.getY());
  }
  gfx.setStroke(new BasicStroke(5));
  gfx.setColor(new Color(0x00ff00));
  gfx.draw(poly);
}

Node.js

我們使用 node-canvas 程式庫在圖片上繪圖。

async function highlightFaces(inputFile, faces, outputFile, PImage) {
  // Open the original image
  const stream = fs.createReadStream(inputFile);
  let promise;
  if (inputFile.match(/\.jpg$/)) {
    promise = PImage.decodeJPEGFromStream(stream);
  } else if (inputFile.match(/\.png$/)) {
    promise = PImage.decodePNGFromStream(stream);
  } else {
    throw new Error(`Unknown filename extension ${inputFile}`);
  }
  const img = await promise;
  const context = img.getContext('2d');
  context.drawImage(img, 0, 0, img.width, img.height, 0, 0);

  // Now draw boxes around all the faces
  context.strokeStyle = 'rgba(0,255,0,0.8)';
  context.lineWidth = '5';

  faces.forEach(face => {
    context.beginPath();
    let origX = 0;
    let origY = 0;
    face.boundingPoly.vertices.forEach((bounds, i) => {
      if (i === 0) {
        origX = bounds.x;
        origY = bounds.y;
        context.moveTo(bounds.x, bounds.y);
      } else {
        context.lineTo(bounds.x, bounds.y);
      }
    });
    context.lineTo(origX, origY);
    context.stroke();
  });

  // Write the result to a file
  console.log(`Writing to file ${outputFile}`);
  const writeStream = fs.createWriteStream(outputFile);
  await PImage.encodePNGToStream(img, writeStream);
}

Python

def highlight_faces(image, faces, output_filename):
    """Draws a polygon around the faces, then saves to output_filename.

    Args:
      image: a file containing the image with the faces.
      faces: a list of faces found in the file. This should be in the format
          returned by the Vision API.
      output_filename: the name of the image file to be created, where the
          faces have polygons drawn around them.
    """
    im = Image.open(image)
    draw = ImageDraw.Draw(im)
    # Sepecify the font-family and the font-size
    for face in faces:
        box = [(vertex.x, vertex.y) for vertex in face.bounding_poly.vertices]
        draw.line(box + [box[0]], width=5, fill="#00ff00")
        # Place the confidence value/score of the detected faces above the
        # detection box in the output image
        draw.text(
            (
                (face.bounding_poly.vertices)[0].x,
                (face.bounding_poly.vertices)[0].y - 30,
            ),
            str(format(face.detection_confidence, ".3f")) + "%",
            fill="#FF0000",
        )
    im.save(output_filename)

馬上開始全面整合吧！

Java

/** Annotates an image using the Vision API. */
public static void main(String[] args) throws IOException, GeneralSecurityException {
  if (args.length != 2) {
    System.err.println("Usage:");
    System.err.printf(
        "\tjava %s inputImagePath outputImagePath\n", FaceDetectApp.class.getCanonicalName());
    System.exit(1);
  }
  Path inputPath = Paths.get(args[0]);
  Path outputPath = Paths.get(args[1]);
  if (!outputPath.toString().toLowerCase().endsWith(".jpg")) {
    System.err.println("outputImagePath must have the file extension 'jpg'.");
    System.exit(1);
  }

  FaceDetectApp app = new FaceDetectApp(getVisionService());
  List<FaceAnnotation> faces = app.detectFaces(inputPath, MAX_RESULTS);
  System.out.printf("Found %d face%s\n", faces.size(), faces.size() == 1 ? "" : "s");
  System.out.printf("Writing to file %s\n", outputPath);
  app.writeWithFaces(inputPath, outputPath, faces);
}

...

如要建構及執行範例，請從範例程式碼目錄執行下列指令：

mvn clean compile assembly:single
java -cp target/vision-face-detection-1.0-SNAPSHOT-jar-with-dependencies.jar \
    com.google.cloud.vision.samples.facedetect.FaceDetectApp \
    data/face.jpg \
    output.jpg

Node.js

async function main(inputFile, outputFile) {
  const PImage = require('pureimage');
  outputFile = outputFile || 'out.png';
  const faces = await detectFaces(inputFile);
  console.log('Highlighting...');
  await highlightFaces(inputFile, faces, outputFile, PImage);
  console.log('Finished!');
}

如要執行程式碼範例，請從程式碼範例目錄執行下列指令：

node faceDetection resources/face.png

Python

def main(input_filename, output_filename, max_results):
    with open(input_filename, "rb") as image:
        faces = detect_face(image, max_results)
        print("Found {} face{}".format(len(faces), "" if len(faces) == 1 else "s"))

        print(f"Writing to file {output_filename}")
        # Reset the file pointer, so we can read the file again
        image.seek(0)
        highlight_faces(image, faces, output_filename)

清除所用資源

為避免因為本教學課程所用資源，導致系統向 Google Cloud 帳戶收取費用，請刪除含有相關資源的專案，或者保留專案但刪除個別資源。

注意：刪除專案會造成以下結果：

• 專案中的所有內容都會遭到刪除。如果使用現有專案來進行本文中的任務，刪除專案將一併移除當中已完成的其他任務'。
• 自訂專案 ID 會消失。當您之前建立這個專案時，可能建立了想要在日後使用的自訂專案 ID。如要保留使用該專案 ID 的網址 (例如 appspot.com 網址)，請刪除在該專案中選取的資源，而不是刪除整個專案。

如果打算探索多種架構、教學課程或快速入門導覽課程，重複使用專案可避免超出專案配額限制。

1. 前往 Google Cloud 控制台的「Manage resources」(管理資源) 頁面。

   前往「Manage resources」(管理資源)

2. 在專案清單中選取要刪除的專案，然後點選「Delete」(刪除)。
3. 在對話方塊中輸入專案 ID，然後按一下 [Shut down] (關閉) 以刪除專案。