Tutorial de detección de caras

Crear el objeto de servicio

Para acceder a las APIs de Google mediante los SDKs de cliente oficiales, debes crear un objeto de servicio basado en el documento de descubrimiento de la API, que describe la API al SDK. Deberás obtenerlo del servicio de descubrimiento de la API Vision con tus credenciales:

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()

Enviar una solicitud de detección de caras

Para crear una solicitud a la API Vision, primero consulta la documentación de la API. En este caso, le pedirás al recurso images que annotate tu imagen. Una solicitud a esta API tiene la forma de un objeto con una lista requests. Cada elemento de esta lista contiene dos datos:

• Los datos de imagen codificados en Base64
• Una lista de las funciones que quieres que se anoten sobre esa imagen.

En este ejemplo, solo solicitarás la anotación FACE_DETECTION en una imagen y devolverás la parte pertinente de la respuesta:

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

Procesar la respuesta

¡Enhorabuena! Has detectado las caras de tu imagen. La respuesta a nuestra solicitud de anotación de caras incluye un conjunto de metadatos sobre las caras detectadas, como las coordenadas de un polígono que abarca la cara. Sin embargo, en este punto, solo se trata de una lista de números. Vamos a usarlas para confirmar que has encontrado las caras de la imagen. Dibujaremos polígonos en una copia de la imagen con las coordenadas devueltas por la API Vision:

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

Usamos la biblioteca node-canvas para dibujar en imágenes.

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)

Analizar todos los datos en conjunto

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);
}

...

Para compilar y ejecutar el ejemplo, ejecuta los siguientes comandos desde el directorio del código de ejemplo:

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!');
}

Para ejecutar el ejemplo, ejecuta el siguiente comando desde el directorio del código de ejemplo:

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)