在线小批量文件注释

Vision API 可为 Cloud Storage 中存储的 PDF、TIFF 或 GIF 文件中的多个页面或帧提供在线(即时)注释。

您可以请求对为每个文件选择的 5 个帧(GIF;“image/gif”)或页面(PDF;“application/pdf”或 TIFF;“image/tiff”)进行在线特征检测和注释。

此页面中的示例注释针对的是 DOCUMENT_TEXT_DETECTION,而在线小批量注释适用于所有 Vision 特征

PDF 文件的前五页
gs://cloud-samples-data/vision/document_understanding/custom_0773375000.pdf

第 1 页

示例 PDF 的第 1 页

...
"text": "á\n7.1.15\nOIL, GAS AND MINERAL LEASE
\nNORVEL J. CHITTIM, ET AL\n.\n.
\nTO\nW. L. SCHEIG\n"
},
"context": {"pageNumber": 1}
...

第 2 页

示例 PDF 的第 2 页(上)

...
"text": "...\n.\n*\n.\n.\n.\nA\nNY\nALA...\n7
\n| THE STATE OF TEXAS
\nOIL, GAS AND MINERAL LEASE
\nCOUNTY OF MAVERICK ]
\nTHIS AGREEMENT made this 14 day of_June
\n1954, between Norvel J. Chittim and his wife, Lieschen G. Chittim;
\nMary Anne Chittim Parker, joined herein pro forma by her husband,
\nJoseph Bright Parker; Dorothea Chittim Oppenheimer, joined herein
\npro forma by her husband, Fred J. Oppenheimer; Tuleta Chittim
\nWright, joined herein pro forma by her husband, Gilbert G. Wright,
\nJr.; Gilbert G. Wright, III; Dela Wright White, joined herein pro
\nforma by her husband, John H. White; Anne Wright Basse, joined
\nherein pro forma by her husband, E. A. Basse, Jr.; Norvel J.
\nChittim, Independent Executor and Trustee for Estate of Marstella
\nChittim, Deceased; Mary Louise Roswell, joined herein pro forma by
\nher husband, Charles M. 'Roswell; and James M. Chittim and his wife,
\nThelma Neal Chittim; as LESSORS, and W. L. Scheig of San Antonio,
\nTexas, as LESSEE,
示例 PDF 的第 2 页(下)

\nW I T N E s s E T H:
\n1. Lessors, in consideration of $10.00, cash in hand paid,
\nof the royalties herein provided, and of the agreement of Lessee
\nherein contained, hereby grant, lease and let exclusively unto
\nLessee the tracts of land hereinafter described for the purpose of
\ntesting for mineral indications, and in such tests use the Seismo-
\ngraph, Torsion Balance, Core Drill, or any other tools, machinery,
\nequipment or explosive necessary and proper; and also prospecting,
\ndrilling and mining for and producing oil, gas and other minerals
\n(except metallic minerals), laying pipe lines, building tanks,
\npower stations, telephone lines and other structures thereon to
\nproduce, save, take care of, treat, transport and own said pro-
\nducts and housing its employees (Lessee to conduct its geophysical
\nwork in such manner as not to damage the buildings, water tanks
\nor wells of Lessors, or the livestock of Lessors or Lessors' ten- !
\nants, )said lands being situated in Maverick, Zavalla and Dimmit
\nCounties, Texas, to-wit:\n3-1.\n"
},
"context": {"pageNumber": 2}
...

第 3 页

示例 PDF 的第 3 页(上)

...
"text": "Being a tract consisting of 140,769.86 acres, more or
\nless, out of what is known as the \"Chittim Ranch\" in said counties,
\nas designated and described in Exhibit \"A\" hereto attached and
\nmade a part hereof as if fully written herein. It being under-
\nstood that the acreage intended to be included in this lease aggre-
\ngates approximately 140,769.86 acres whether it actually comprises
\nmore or less, but for the purpose of calculating the payments
\nhereinafter provided for, it is agreed that the land included with-
\nin the terms of this lease is One hundred forty thousand seven
\nhundred sixty-nine and eighty-six one hundredths (140,769.86) acres,
\nand that each survey listed above contains the acreage stated above.
\nIt is understood that tract designated \"TRACT II\" in
\nExhibit \"A\" is subject to a one-sixteenth (1/16) royalty reserved.
\nto the State of Texas, and the rights of the State of Texas must
\nbe respected in the development of the said property.
示例 PDF 的第 3 页(下)

\n2. Subject to the other provisions hereof, this lease shall
\nbe for a term of ten (10) years from date hereof (called \"Primary
\nTerm\"), and as long thereafter as oil, gas or other minerals
\n(except metallic minerals) are produced from said land hereunder
\nin paying quantities, subject, however, to all of the terms and
\nprovisions of this lease. After expiration of the primary term,
\nthis lease shall terminate as to all lands included herein, save
\nand except as to those tracts which lessee maintains in force and
\neffect according to the requirements hereof.
\n3. The royalties to be paid by Lessee are (a) on oil, one-
\neighth (1/8) of that produced and saved from said land, the same to
\nbe delivered at the well or to the credit of Lessors into the pipe i
\nline to which the well may be connected; (b) on gas, including
\ni casinghead gas or other gaseous or vaporous substance, produced
\nfrom the leased premises and sold or used by Lessee off the leased
\npremises or in the manufacture of gasoline or other products, the
\nmarket value, at the mouth of the well, of one-eighth (1/8) of
\n.\n3-2-\n?\n"
},
"context": {"pageNumber": 3}
...

第 4 页

示例 PDF 的第 4 页(上)

...
"text": "•\n:\n.\nthe gas or casinghead gas so used or sold. On all gas or casing-
\nhead gas sold at the well, the royalty shall be one-eighth (1/8)
\nof the amounts realized from such sales. While gas from any well
\nproducing gas only is being used or sold by. Lessee, Lessor may have
\nenough of said gas for all stoves and inside lights in the prin-
\ncipal dwelling house on the leased premises by making Lessors' own
\nconnections with the well and by assuming all risk and paying all
\nexpenses. And (c) on all other minerals (except metallic minerals)
\nmined and marketed, one tenth (1/10). either in kind or value at the
\nwell or mine at Lessee's election.
\nFor the purpose of royalty payments under 3 (b) hereof,
\nall liquid hydrocarbons (including distillate) recovered and saved
n| by Lessee in separators or traps on the leased premises shall be
\nconsidered as oil. Should such a plant be constructed by another
\nthan Lessee to whom Lessee should sell or deliver the gas or cas-
\ninghead gas produced from the leased premises for processing, then
\nthe royalty thereon shall be one-eighth (1/8) of the amounts
\nrealized by Lessee from such sales or deliveries.
示例 PDF 的第 4 页(下)

\nOr if such plant is owned or constructed or operated by
\nLessee, then the royalty shall be on the basis of one-eighth (1/8) |
\nof the prevailing price in the area for such products..
\nThe provisions of this paragraph shall control as to any
\nconflict with Paragraph 3 (b). Lessors shall also be entitled to
\nsaid royalty interest in all residue gas .obtained, saved and mar-
\nketed from said premises, or used off the premises, or that may be
\nreplaced in the reservoir by 'any recycling process, settlement
\ntherefor to be made to Lessors when such gas is marketed or used
\noff the premises. !
\nIf at the expiration of the primary term of this lease
\nLessee has not found and produced oil or gas in paying quantities
\nin any formation lying fifty (50) feet below the base of what is
\nknown as the Rhodessa section at the particular point where the
\nwell is drilled, then, subject to the further provisions hereof,
\nthis lease shall terminate as to all horizons below fifty (50)
\nI feet below the Rhodessa section. And if at the expiration of the
\n3 -3-\n"
},
"context": {"pageNumber": 4}
...

第 5 页

示例 PDF 的第 5 页(上)

...
"text": ".\n.\n:\nI\n.\n.\n.:250:-....\n.\n...\n.\n....\n....\n..\n..\n. ..
\n.\n..\n.\n...\n...\n.-\n.\n.\n..\n..\n17\n.\n:\n-\n-\n-\n.\n..\n.
\nprimary term production of oil or gas in paying quantities is not
\nfound in the Jurassic, then this lease shall terminate as to the
\nJurassic and lower formations unless Lessee shall have completed
\nat least two (2) tests in the Jurassic. And after the primary
\nterm Lessee shall complete at least one (1) Jurassic test each
\nthree years on said property as to which this lease is still in
\neffect, until paying production is obtained in or below the
\nJurassic, or upon failure so to do Lessee shall release this
\nlease as to all formations below the top of the Jurassic. Upon
\ncompliance with the above provisions as to Jurassic tests, and
\nif production is found in the Jurassic, then, subject to the
\nother provisions hereof, this lease shall be effective as to all
\nhorizons, including the Jurassic..
\n5. It is understood and expressly agreed that the consider-
\niation first recited in this lease, the down cash payment, receipt
\nof which is hereby acknowledged by Lessors, is full and adequate
\nconsideration to maintain this lease in full force and effect for
\na period of one year from the date hereof, and does not impose
\nany obligation on the part of Lessee to drill and develop this
\nlease during the said term of one year from date of this lease.
示例 PDF 的第 5 页(下)

\n6. This lease shall terminate as to both parties unless
\non or before one year from this date, Lessee shall pay to or ten- !
\nder to Lessors or to the credit of Lessors, in the National Bank
\nof Commerce, at San Antonio, Texas, (which bank and its successors
\nare Lessors' agent, and shall continue as the depository for all \"
\nrental payable hereunder regardless of changes in ownership of
\nsaid land or the rental), the sum of One Dollar ($1.00) per acre
\nas to all acreage then covered by this lease, and not surrendered,
\nor maintained by production of oil, gas or other minerals, or by
\ndrilling-reworking operations, all as hereinafter fully set out, :
\nwhich shall maintain this lease in full force and effect for
\nanother twelve-month period, without imposing any obligation on
\nthe part of Lessee to drill and develop this lease. In like
\nmanner, and upon like payment or tender annually, Lessee may
\nmaintain this lease .in full force and effect for successive
\ntwelve-month periods during the primary term, without imposing
\n.\n--.\n.\n.\n.\n-\n::\n---
\n-\n3\n.\n..-\n-\n-\n:.\n.\n::\n.
\n3-4-\n"
},
"context": {"pageNumber": 5}
...

限制

最多可以为 5 个页面添加注释。用户可以指定 5 个特定页面来添加注释。

身份验证

设置您的 Google Cloud 项目和身份验证

目前支持的特征类型

特征类型
CROP_HINTS 确定图片的建议剪裁区域顶点。
DOCUMENT_TEXT_DETECTION 对文档 (PDF/TIFF) 等包含密集文本的图片和包含手写内容的图片执行 OCR。TEXT_DETECTION 可用于包含稀疏文本的图片。 如果同时存在 DOCUMENT_TEXT_DETECTIONTEXT_DETECTION,则优先考虑。
FACE_DETECTION 检测图片中的人脸。
IMAGE_PROPERTIES 计算一组图片属性,例如图片的主色。
LABEL_DETECTION 根据图片内容添加标签。
LANDMARK_DETECTION 检测图片中的地标。
LOGO_DETECTION 检测图片中的公司徽标。
OBJECT_LOCALIZATION 检测并提取图片中的多个对象。
SAFE_SEARCH_DETECTION 运行安全搜索可检测可能不安全的内容或不良内容。
TEXT_DETECTION 对图片中的文本执行光学字符识别 (OCR)。 文本检测针对大型图片中的稀疏文本区域进行了优化。 如果图片为文档 (PDF/TIFF)、包含密集文本或包含手写内容,请改用 DOCUMENT_TEXT_DETECTION
WEB_DETECTION 检测图片中的新闻、事件或名人等主题实体,并借助强大的 Google 图片搜索在网络上查找相似的图片。

示例代码

您可以使用本地存储的文件发送注释请求,也可以使用 Cloud Storage 上存储的文件

使用本地存储的文件

使用以下代码示例获取本地存储的文件的任何特征注释。

REST

如需对一小批文件执行在线 PDF/TIFF/GIF 特征检测,请发出 POST 请求并提供相应的请求正文:

在使用任何请求数据之前,请先进行以下替换:

  • BASE64_ENCODED_FILE:二进制文件数据的 base64 表示(ASCII 字符串)。此字符串应类似于以下字符串:
    • JVBERi0xLjUNCiW1tbW1...ydHhyZWYNCjk5NzM2OQ0KJSVFT0Y=
    如需了解详情,请参阅 base64 编码主题。
  • PROJECT_ID:您的 Google Cloud 项目 ID。

特定于字段的注意事项

  • inputConfig.mimeType - 下列类型之一:“application/pdf”“image/tiff”或“image/gif”。
  • pages - 指定要执行特征检测的文件的特定页面。

HTTP 方法和网址:

POST https://vision.googleapis.com/v1/files:annotate

请求 JSON 正文:

{
  "requests": [
    {
      "inputConfig": {
        "content": "BASE64_ENCODED_FILE",
        "mimeType": "application/pdf"
      },
      "features": [
        {
          "type": "DOCUMENT_TEXT_DETECTION"
        }
      ],
      "pages": [
        1,2,3,4,5
      ]
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中,然后执行以下命令:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "x-goog-user-project: PROJECT_ID" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://vision.googleapis.com/v1/files:annotate"

PowerShell

将请求正文保存在名为 request.json 的文件中,然后执行以下命令:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://vision.googleapis.com/v1/files:annotate" | Select-Object -Expand Content
响应:

成功的 annotate 请求会立即返回 JSON 响应。

对于此特征 (DOCUMENT_TEXT_DETECTION),JSON 响应与图片的文档文本检测请求的响应类似。响应包含按段落、字词和各符号划分的文本块的边界框。还会检测全文。响应还包含一个 context 字段,显示指定的 PDF 或 TIFF 的位置以及结果在文件中的页码。

以下响应 JSON 仅针对单个页面(第 2 页),为清楚起见,这里采用简写形式。

Java

在试用此示例之前,请按照Vision API 快速入门:使用客户端库中的 Java 设置说明进行操作。如需了解详情,请参阅 Vision API Java 参考文档

import com.google.cloud.vision.v1.AnnotateFileRequest;
import com.google.cloud.vision.v1.AnnotateImageResponse;
import com.google.cloud.vision.v1.BatchAnnotateFilesRequest;
import com.google.cloud.vision.v1.BatchAnnotateFilesResponse;
import com.google.cloud.vision.v1.Block;
import com.google.cloud.vision.v1.Feature;
import com.google.cloud.vision.v1.ImageAnnotatorClient;
import com.google.cloud.vision.v1.InputConfig;
import com.google.cloud.vision.v1.Page;
import com.google.cloud.vision.v1.Paragraph;
import com.google.cloud.vision.v1.Symbol;
import com.google.cloud.vision.v1.Word;
import com.google.protobuf.ByteString;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;

public class BatchAnnotateFiles {

  public static void batchAnnotateFiles() throws IOException {
    String filePath = "path/to/your/file.pdf";
    batchAnnotateFiles(filePath);
  }

  public static void batchAnnotateFiles(String filePath) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (ImageAnnotatorClient imageAnnotatorClient = ImageAnnotatorClient.create()) {
      // You can send multiple files to be annotated, this sample demonstrates how to do this with
      // one file. If you want to use multiple files, you have to create a `AnnotateImageRequest`
      // object for each file that you want annotated.
      // First read the files contents
      Path path = Paths.get(filePath);
      byte[] data = Files.readAllBytes(path);
      ByteString content = ByteString.copyFrom(data);

      // Specify the input config with the file's contents and its type.
      // Supported mime_type: application/pdf, image/tiff, image/gif
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#inputconfig
      InputConfig inputConfig =
          InputConfig.newBuilder().setMimeType("application/pdf").setContent(content).build();

      // Set the type of annotation you want to perform on the file
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.Feature.Type
      Feature feature = Feature.newBuilder().setType(Feature.Type.DOCUMENT_TEXT_DETECTION).build();

      // Build the request object for that one file. Note: for additional file you have to create
      // additional `AnnotateFileRequest` objects and store them in a list to be used below.
      // Since we are sending a file of type `application/pdf`, we can use the `pages` field to
      // specify which pages to process. The service can process up to 5 pages per document file.
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.AnnotateFileRequest
      AnnotateFileRequest fileRequest =
          AnnotateFileRequest.newBuilder()
              .setInputConfig(inputConfig)
              .addFeatures(feature)
              .addPages(1) // Process the first page
              .addPages(2) // Process the second page
              .addPages(-1) // Process the last page
              .build();

      // Add each `AnnotateFileRequest` object to the batch request.
      BatchAnnotateFilesRequest request =
          BatchAnnotateFilesRequest.newBuilder().addRequests(fileRequest).build();

      // Make the synchronous batch request.
      BatchAnnotateFilesResponse response = imageAnnotatorClient.batchAnnotateFiles(request);

      // Process the results, just get the first result, since only one file was sent in this
      // sample.
      for (AnnotateImageResponse imageResponse :
          response.getResponsesList().get(0).getResponsesList()) {
        System.out.format("Full text: %s%n", imageResponse.getFullTextAnnotation().getText());
        for (Page page : imageResponse.getFullTextAnnotation().getPagesList()) {
          for (Block block : page.getBlocksList()) {
            System.out.format("%nBlock confidence: %s%n", block.getConfidence());
            for (Paragraph par : block.getParagraphsList()) {
              System.out.format("\tParagraph confidence: %s%n", par.getConfidence());
              for (Word word : par.getWordsList()) {
                System.out.format("\t\tWord confidence: %s%n", word.getConfidence());
                for (Symbol symbol : word.getSymbolsList()) {
                  System.out.format(
                      "\t\t\tSymbol: %s, (confidence: %s)%n",
                      symbol.getText(), symbol.getConfidence());
                }
              }
            }
          }
        }
      }
    }
  }
}

Node.js

试用此示例之前,请按照《Vision 快速入门:使用客户端库》中的 Node.js 设置说明进行操作。 如需了解详情,请参阅 Vision Node.js API 参考文档

如需向 Vision 进行身份验证,请设置应用默认凭据。 如需了解详情,请参阅为本地开发环境设置身份验证

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const fileName = 'path/to/your/file.pdf';

// Imports the Google Cloud client libraries
const {ImageAnnotatorClient} = require('@google-cloud/vision').v1;
const fs = require('fs').promises;

// Instantiates a client
const client = new ImageAnnotatorClient();

// You can send multiple files to be annotated, this sample demonstrates how to do this with
// one file. If you want to use multiple files, you have to create a request object for each file that you want annotated.
async function batchAnnotateFiles() {
  // First Specify the input config with the file's path and its type.
  // Supported mime_type: application/pdf, image/tiff, image/gif
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#inputconfig
  const inputConfig = {
    mimeType: 'application/pdf',
    content: await fs.readFile(fileName),
  };

  // Set the type of annotation you want to perform on the file
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.Feature.Type
  const features = [{type: 'DOCUMENT_TEXT_DETECTION'}];

  // Build the request object for that one file. Note: for additional files you have to create
  // additional file request objects and store them in a list to be used below.
  // Since we are sending a file of type `application/pdf`, we can use the `pages` field to
  // specify which pages to process. The service can process up to 5 pages per document file.
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.AnnotateFileRequest
  const fileRequest = {
    inputConfig: inputConfig,
    features: features,
    // Annotate the first two pages and the last one (max 5 pages)
    // First page starts at 1, and not 0. Last page is -1.
    pages: [1, 2, -1],
  };

  // Add each `AnnotateFileRequest` object to the batch request.
  const request = {
    requests: [fileRequest],
  };

  // Make the synchronous batch request.
  const [result] = await client.batchAnnotateFiles(request);

  // Process the results, just get the first result, since only one file was sent in this
  // sample.
  const responses = result.responses[0].responses;

  for (const response of responses) {
    console.log(`Full text: ${response.fullTextAnnotation.text}`);
    for (const page of response.fullTextAnnotation.pages) {
      for (const block of page.blocks) {
        console.log(`Block confidence: ${block.confidence}`);
        for (const paragraph of block.paragraphs) {
          console.log(` Paragraph confidence: ${paragraph.confidence}`);
          for (const word of paragraph.words) {
            const symbol_texts = word.symbols.map(symbol => symbol.text);
            const word_text = symbol_texts.join('');
            console.log(
              `  Word text: ${word_text} (confidence: ${word.confidence})`
            );
            for (const symbol of word.symbols) {
              console.log(
                `   Symbol: ${symbol.text} (confidence: ${symbol.confidence})`
              );
            }
          }
        }
      }
    }
  }
}

batchAnnotateFiles();

Python

试用此示例之前,请按照《Vision 快速入门:使用客户端库》中的 Python 设置说明进行操作。 如需了解详情,请参阅 Vision Python API 参考文档

如需向 Vision 进行身份验证,请设置应用默认凭据。 如需了解详情,请参阅为本地开发环境设置身份验证



from google.cloud import vision_v1


def sample_batch_annotate_files(file_path="path/to/your/document.pdf"):
    """Perform batch file annotation."""
    client = vision_v1.ImageAnnotatorClient()

    # Supported mime_type: application/pdf, image/tiff, image/gif
    mime_type = "application/pdf"
    with open(file_path, "rb") as f:
        content = f.read()
    input_config = {"mime_type": mime_type, "content": content}
    features = [{"type_": vision_v1.Feature.Type.DOCUMENT_TEXT_DETECTION}]

    # The service can process up to 5 pages per document file. Here we specify
    # the first, second, and last page of the document to be processed.
    pages = [1, 2, -1]
    requests = [{"input_config": input_config, "features": features, "pages": pages}]

    response = client.batch_annotate_files(requests=requests)
    for image_response in response.responses[0].responses:
        print(f"Full text: {image_response.full_text_annotation.text}")
        for page in image_response.full_text_annotation.pages:
            for block in page.blocks:
                print(f"\nBlock confidence: {block.confidence}")
                for par in block.paragraphs:
                    print(f"\tParagraph confidence: {par.confidence}")
                    for word in par.words:
                        print(f"\t\tWord confidence: {word.confidence}")
                        for symbol in word.symbols:
                            print(
                                "\t\t\tSymbol: {}, (confidence: {})".format(
                                    symbol.text, symbol.confidence
                                )
                            )

使用 Cloud Storage 上的文件

使用以下代码示例获取 Cloud Storage 文件的任何特征注释。

REST

如需对一小批文件执行在线 PDF/TIFF/GIF 特征检测,请发出 POST 请求并提供相应的请求正文:

在使用任何请求数据之前,请先进行以下替换:

  • CLOUD_STORAGE_FILE_URI:Cloud Storage 存储桶中有效文件 (PDF/TIFF) 的路径。您必须至少拥有该文件的读取权限。 示例:
    • gs://cloud-samples-data/vision/document_understanding/custom_0773375000.pdf
  • PROJECT_ID:您的 Google Cloud 项目 ID。

特定于字段的注意事项

  • inputConfig.mimeType - 下列类型之一:“application/pdf”“image/tiff”或“image/gif”。
  • pages - 指定要执行特征检测的文件的特定页面。

HTTP 方法和网址:

POST https://vision.googleapis.com/v1/files:annotate

请求 JSON 正文:

{
  "requests": [
    {
      "inputConfig": {
        "gcsSource": {
          "uri": "CLOUD_STORAGE_FILE_URI"
        },
        "mimeType": "application/pdf"
      },
      "features": [
        {
          "type": "DOCUMENT_TEXT_DETECTION"
        }
      ],
      "pages": [
        1,2,3,4,5
      ]
    }
  ]
}

如需发送请求,请选择以下方式之一:

curl

将请求正文保存在名为 request.json 的文件中,然后执行以下命令:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "x-goog-user-project: PROJECT_ID" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://vision.googleapis.com/v1/files:annotate"

PowerShell

将请求正文保存在名为 request.json 的文件中,然后执行以下命令:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred"; "x-goog-user-project" = "PROJECT_ID" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://vision.googleapis.com/v1/files:annotate" | Select-Object -Expand Content
响应:

成功的 annotate 请求会立即返回 JSON 响应。

对于此特征 (DOCUMENT_TEXT_DETECTION),JSON 响应与图片的文档文本检测请求的响应类似。响应包含按段落、字词和各符号划分的文本块的边界框。还会检测全文。响应还包含一个 context 字段,显示指定的 PDF 或 TIFF 的位置以及结果在文件中的页码。

以下响应 JSON 仅针对单个页面(第 2 页),为清楚起见,这里采用简写形式。

Java

在试用此示例之前,请按照Vision API 快速入门:使用客户端库中的 Java 设置说明进行操作。如需了解详情,请参阅 Vision API Java 参考文档

import com.google.cloud.vision.v1.AnnotateFileRequest;
import com.google.cloud.vision.v1.AnnotateImageResponse;
import com.google.cloud.vision.v1.BatchAnnotateFilesRequest;
import com.google.cloud.vision.v1.BatchAnnotateFilesResponse;
import com.google.cloud.vision.v1.Block;
import com.google.cloud.vision.v1.Feature;
import com.google.cloud.vision.v1.GcsSource;
import com.google.cloud.vision.v1.ImageAnnotatorClient;
import com.google.cloud.vision.v1.InputConfig;
import com.google.cloud.vision.v1.Page;
import com.google.cloud.vision.v1.Paragraph;
import com.google.cloud.vision.v1.Symbol;
import com.google.cloud.vision.v1.Word;
import java.io.IOException;

public class BatchAnnotateFilesGcs {

  public static void batchAnnotateFilesGcs() throws IOException {
    String gcsUri = "gs://cloud-samples-data/vision/document_understanding/kafka.pdf";
    batchAnnotateFilesGcs(gcsUri);
  }

  public static void batchAnnotateFilesGcs(String gcsUri) throws IOException {
    // Initialize client that will be used to send requests. This client only needs to be created
    // once, and can be reused for multiple requests. After completing all of your requests, call
    // the "close" method on the client to safely clean up any remaining background resources.
    try (ImageAnnotatorClient imageAnnotatorClient = ImageAnnotatorClient.create()) {
      // You can send multiple files to be annotated, this sample demonstrates how to do this with
      // one file. If you want to use multiple files, you have to create a `AnnotateImageRequest`
      // object for each file that you want annotated.
      // First specify where the vision api can find the image
      GcsSource gcsSource = GcsSource.newBuilder().setUri(gcsUri).build();

      // Specify the input config with the file's uri and its type.
      // Supported mime_type: application/pdf, image/tiff, image/gif
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#inputconfig
      InputConfig inputConfig =
          InputConfig.newBuilder().setMimeType("application/pdf").setGcsSource(gcsSource).build();

      // Set the type of annotation you want to perform on the file
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.Feature.Type
      Feature feature = Feature.newBuilder().setType(Feature.Type.DOCUMENT_TEXT_DETECTION).build();

      // Build the request object for that one file. Note: for additional file you have to create
      // additional `AnnotateFileRequest` objects and store them in a list to be used below.
      // Since we are sending a file of type `application/pdf`, we can use the `pages` field to
      // specify which pages to process. The service can process up to 5 pages per document file.
      // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.AnnotateFileRequest
      AnnotateFileRequest fileRequest =
          AnnotateFileRequest.newBuilder()
              .setInputConfig(inputConfig)
              .addFeatures(feature)
              .addPages(1) // Process the first page
              .addPages(2) // Process the second page
              .addPages(-1) // Process the last page
              .build();

      // Add each `AnnotateFileRequest` object to the batch request.
      BatchAnnotateFilesRequest request =
          BatchAnnotateFilesRequest.newBuilder().addRequests(fileRequest).build();

      // Make the synchronous batch request.
      BatchAnnotateFilesResponse response = imageAnnotatorClient.batchAnnotateFiles(request);

      // Process the results, just get the first result, since only one file was sent in this
      // sample.
      for (AnnotateImageResponse imageResponse :
          response.getResponsesList().get(0).getResponsesList()) {
        System.out.format("Full text: %s%n", imageResponse.getFullTextAnnotation().getText());
        for (Page page : imageResponse.getFullTextAnnotation().getPagesList()) {
          for (Block block : page.getBlocksList()) {
            System.out.format("%nBlock confidence: %s%n", block.getConfidence());
            for (Paragraph par : block.getParagraphsList()) {
              System.out.format("\tParagraph confidence: %s%n", par.getConfidence());
              for (Word word : par.getWordsList()) {
                System.out.format("\t\tWord confidence: %s%n", word.getConfidence());
                for (Symbol symbol : word.getSymbolsList()) {
                  System.out.format(
                      "\t\t\tSymbol: %s, (confidence: %s)%n",
                      symbol.getText(), symbol.getConfidence());
                }
              }
            }
          }
        }
      }
    }
  }
}

Node.js

试用此示例之前,请按照《Vision 快速入门:使用客户端库》中的 Node.js 设置说明进行操作。 如需了解详情,请参阅 Vision Node.js API 参考文档

如需向 Vision 进行身份验证,请设置应用默认凭据。 如需了解详情,请参阅为本地开发环境设置身份验证

/**
 * TODO(developer): Uncomment these variables before running the sample.
 */
// const gcsSourceUri = 'gs://cloud-samples-data/vision/document_understanding/kafka.pdf';

// Imports the Google Cloud client libraries
const {ImageAnnotatorClient} = require('@google-cloud/vision').v1;

// Instantiates a client
const client = new ImageAnnotatorClient();

// You can send multiple files to be annotated, this sample demonstrates how to do this with
// one file. If you want to use multiple files, you have to create a request object for each file that you want annotated.
async function batchAnnotateFiles() {
  // First Specify the input config with the file's uri and its type.
  // Supported mime_type: application/pdf, image/tiff, image/gif
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#inputconfig
  const inputConfig = {
    mimeType: 'application/pdf',
    gcsSource: {
      uri: gcsSourceUri,
    },
  };

  // Set the type of annotation you want to perform on the file
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.Feature.Type
  const features = [{type: 'DOCUMENT_TEXT_DETECTION'}];

  // Build the request object for that one file. Note: for additional files you have to create
  // additional file request objects and store them in a list to be used below.
  // Since we are sending a file of type `application/pdf`, we can use the `pages` field to
  // specify which pages to process. The service can process up to 5 pages per document file.
  // https://cloud.google.com/vision/docs/reference/rpc/google.cloud.vision.v1#google.cloud.vision.v1.AnnotateFileRequest
  const fileRequest = {
    inputConfig: inputConfig,
    features: features,
    // Annotate the first two pages and the last one (max 5 pages)
    // First page starts at 1, and not 0. Last page is -1.
    pages: [1, 2, -1],
  };

  // Add each `AnnotateFileRequest` object to the batch request.
  const request = {
    requests: [fileRequest],
  };

  // Make the synchronous batch request.
  const [result] = await client.batchAnnotateFiles(request);

  // Process the results, just get the first result, since only one file was sent in this
  // sample.
  const responses = result.responses[0].responses;

  for (const response of responses) {
    console.log(`Full text: ${response.fullTextAnnotation.text}`);
    for (const page of response.fullTextAnnotation.pages) {
      for (const block of page.blocks) {
        console.log(`Block confidence: ${block.confidence}`);
        for (const paragraph of block.paragraphs) {
          console.log(` Paragraph confidence: ${paragraph.confidence}`);
          for (const word of paragraph.words) {
            const symbol_texts = word.symbols.map(symbol => symbol.text);
            const word_text = symbol_texts.join('');
            console.log(
              `  Word text: ${word_text} (confidence: ${word.confidence})`
            );
            for (const symbol of word.symbols) {
              console.log(
                `   Symbol: ${symbol.text} (confidence: ${symbol.confidence})`
              );
            }
          }
        }
      }
    }
  }
}

batchAnnotateFiles();

Python

试用此示例之前,请按照《Vision 快速入门:使用客户端库》中的 Python 设置说明进行操作。 如需了解详情,请参阅 Vision Python API 参考文档

如需向 Vision 进行身份验证,请设置应用默认凭据。 如需了解详情,请参阅为本地开发环境设置身份验证


from google.cloud import vision_v1


def sample_batch_annotate_files(
    storage_uri="gs://cloud-samples-data/vision/document_understanding/kafka.pdf",
):
    """Perform batch file annotation."""
    mime_type = "application/pdf"

    client = vision_v1.ImageAnnotatorClient()

    gcs_source = {"uri": storage_uri}
    input_config = {"gcs_source": gcs_source, "mime_type": mime_type}
    features = [{"type_": vision_v1.Feature.Type.DOCUMENT_TEXT_DETECTION}]

    # The service can process up to 5 pages per document file.
    # Here we specify the first, second, and last page of the document to be
    # processed.
    pages = [1, 2, -1]
    requests = [{"input_config": input_config, "features": features, "pages": pages}]

    response = client.batch_annotate_files(requests=requests)
    for image_response in response.responses[0].responses:
        print(f"Full text: {image_response.full_text_annotation.text}")
        for page in image_response.full_text_annotation.pages:
            for block in page.blocks:
                print(f"\nBlock confidence: {block.confidence}")
                for par in block.paragraphs:
                    print(f"\tParagraph confidence: {par.confidence}")
                    for word in par.words:
                        print(f"\t\tWord confidence: {word.confidence}")
                        for symbol in word.symbols:
                            print(
                                "\t\t\tSymbol: {}, (confidence: {})".format(
                                    symbol.text, symbol.confidence
                                )
                            )

试用

请尝试下面的小批量在线特征检测。

您可以使用已指定的 PDF 文件,也可以指定自己的文件。

PDF 文件的前五页
gs://cloud-samples-data/vision/document_understanding/custom_0773375000.pdf

对于此请求,指定了三种特征类型:

  • DOCUMENT_TEXT_DETECTION
  • LABEL_DETECTION
  • CROP_HINTS

您可以通过更改请求中的相应对象 ({"type": "FEATURE_NAME"}) 来添加或移除其他特征类型

选择执行即可发送请求。

请求正文:

{
  "requests": [
    {
      "inputConfig": {
        "gcsSource": {
          "uri": "gs://cloud-samples-data/vision/document_understanding/custom_0773375000.pdf"
        },
        "mimeType": "application/pdf"
      },
      "features": [
        {
          "type": "DOCUMENT_TEXT_DETECTION"
        },
        {
          "type": "LABEL_DETECTION"
        },
        {
          "type": "CROP_HINTS"
        }
      ],
      "pages": [
        1,
        2,
        3,
        4,
        5
      ]
    }
  ]
}