Create a dataset for training video object tracking models

This page shows you how to create a Vertex AI dataset from your video data so you can start training object tracking models. You can create a dataset using either the Google Cloud console or the Vertex AI API.

Create an empty dataset and import or associate your data

Google Cloud console

Use the following instructions to create an empty dataset and either import or associate your data.

  1. In the Google Cloud console, in the Vertex AI section, go to the Datasets page.

    Go to the Datasets page

  2. Click Create to open the create dataset details page.
  3. Modify the Dataset name field to create a descriptive dataset display name.
  4. Select the Video tab.
  5. Select Video object tracking.
  6. Select a region from the Region drop-down list.
  7. Click Create to create your empty dataset, and advance to the data import page.
  8. Choose one of the following options from the Select an import method section:

    Upload data from your computer

    1. In the Select an import method section, choose to upload data from your computer.
    2. Click Select files and choose all the local files to upload to a Cloud Storage bucket.
    3. In the Select a Cloud Storage path section click Browse to choose a Cloud Storage bucket location to upload your data to.

    Upload an import file from your computer

    1. Click Upload an import file from your computer.
    2. Click Select files and choose the local import file to upload to a Cloud Storage bucket.
    3. In the Select a Cloud Storage path section click Browse to choose a Cloud Storage bucket location to upload your file to.

    Select an import file from Cloud Storage

    1. Click Select an import file from Cloud Storage.
    2. In the Select a Cloud Storage path section click Browse to choose the import file in Cloud Storage.
  9. Click Continue.

    Data import can take several hours, depending on the size of your data. You can close this tab and return to it later. You will receive an email when your data is imported.

API

In order to create a machine learning model you must first have a representative collection of data to train with. After importing data you can make modifications and start model training.

Create a dataset

Use the following samples to create a dataset for your data.

REST

Before using any of the request data, make the following replacements:

  • LOCATION: Region where the dataset will be stored. This must be a region that supports dataset resources. For example, us-central1. See List of available locations.
  • PROJECT: Your project ID.
  • DATASET_NAME: Name for the dataset.
  • PROJECT_NUMBER: Your project's automatically generated project number.

HTTP method and URL:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/datasets

Request JSON body:

{
  "display_name": "DATASET_NAME",
  "metadata_schema_uri": "gs://google-cloud-aiplatform/schema/dataset/metadata/video_1.0.0.yaml"
}

To send your request, choose one of these options:

curl

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/datasets"

PowerShell

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/LOCATION/datasets" | Select-Object -Expand Content

You should see output similar to the following. You can use the OPERATION_ID in the response to get the status of the operation.

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION/operations/OPERATION_ID",
  "metadata": {
    "@type": "type.googleapis.com/google.cloud.aiplatform.v1.CreateDatasetOperationMetadata",
    "genericMetadata": {
      "createTime": "2020-07-07T21:27:35.964882Z",
      "updateTime": "2020-07-07T21:27:35.964882Z"
    }
  }
}

Terraform

The following sample uses the google_vertex_ai_dataset Terraform resource to create a video dataset named video-dataset.

To learn how to apply or remove a Terraform configuration, see Basic Terraform commands.

resource "google_vertex_ai_dataset" "video_dataset" {
  display_name        = "video-dataset"
  metadata_schema_uri = "gs://google-cloud-aiplatform/schema/dataset/metadata/video_1.0.0.yaml"
  region              = "us-central1"
}

Java

Before trying this sample, follow the Java setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Java API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.


import com.google.api.gax.longrunning.OperationFuture;
import com.google.cloud.aiplatform.v1.CreateDatasetOperationMetadata;
import com.google.cloud.aiplatform.v1.Dataset;
import com.google.cloud.aiplatform.v1.DatasetServiceClient;
import com.google.cloud.aiplatform.v1.DatasetServiceSettings;
import com.google.cloud.aiplatform.v1.LocationName;
import java.io.IOException;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.TimeoutException;

public class CreateDatasetVideoSample {

  public static void main(String[] args)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    // TODO(developer): Replace these variables before running the sample.
    String project = "YOUR_PROJECT_ID";
    String datasetVideoDisplayName = "YOUR_DATASET_VIDEO_DISPLAY_NAME";
    createDatasetSample(datasetVideoDisplayName, project);
  }

  static void createDatasetSample(String datasetVideoDisplayName, String project)
      throws IOException, InterruptedException, ExecutionException, TimeoutException {
    DatasetServiceSettings datasetServiceSettings =
        DatasetServiceSettings.newBuilder()
            .setEndpoint("us-central1-aiplatform.googleapis.com:443")
            .build();

    // 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 (DatasetServiceClient datasetServiceClient =
        DatasetServiceClient.create(datasetServiceSettings)) {
      String location = "us-central1";
      String metadataSchemaUri =
          "gs://google-cloud-aiplatform/schema/dataset/metadata/video_1.0.0.yaml";
      LocationName locationName = LocationName.of(project, location);
      Dataset dataset =
          Dataset.newBuilder()
              .setDisplayName(datasetVideoDisplayName)
              .setMetadataSchemaUri(metadataSchemaUri)
              .build();

      OperationFuture<Dataset, CreateDatasetOperationMetadata> datasetFuture =
          datasetServiceClient.createDatasetAsync(locationName, dataset);
      System.out.format("Operation name: %s\n", datasetFuture.getInitialFuture().get().getName());
      System.out.println("Waiting for operation to finish...");
      Dataset datasetResponse = datasetFuture.get(300, TimeUnit.SECONDS);

      System.out.println("Create Dataset Video Response");
      System.out.format("Name: %s\n", datasetResponse.getName());
      System.out.format("Display Name: %s\n", datasetResponse.getDisplayName());
      System.out.format("Metadata Schema Uri: %s\n", datasetResponse.getMetadataSchemaUri());
      System.out.format("Metadata: %s\n", datasetResponse.getMetadata());
      System.out.format("Create Time: %s\n", datasetResponse.getCreateTime());
      System.out.format("Update Time: %s\n", datasetResponse.getUpdateTime());
      System.out.format("Labels: %s\n", datasetResponse.getLabelsMap());
    }
  }
}

Node.js

Before trying this sample, follow the Node.js setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Node.js API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.

/**
 * TODO(developer): Uncomment these variables before running the sample.\
 * (Not necessary if passing values as arguments)
 */

// const datasetDisplayName = "YOUR_DATASTE_DISPLAY_NAME";
// const project = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION';

// Imports the Google Cloud Dataset Service Client library
const {DatasetServiceClient} = require('@google-cloud/aiplatform');

// Specifies the location of the api endpoint
const clientOptions = {
  apiEndpoint: 'us-central1-aiplatform.googleapis.com',
};

// Instantiates a client
const datasetServiceClient = new DatasetServiceClient(clientOptions);

async function createDatasetVideo() {
  // Configure the parent resource
  const parent = `projects/${project}/locations/${location}`;
  // Configure the dataset resource
  const dataset = {
    displayName: datasetDisplayName,
    metadataSchemaUri:
      'gs://google-cloud-aiplatform/schema/dataset/metadata/video_1.0.0.yaml',
  };
  const request = {
    parent,
    dataset,
  };

  // Create Dataset Request
  const [response] = await datasetServiceClient.createDataset(request);
  console.log(`Long running operation: ${response.name}`);

  // Wait for operation to complete
  await response.promise();
  const result = response.result;

  console.log('Create dataset video response');
  console.log(`Name : ${result.name}`);
  console.log(`Display name : ${result.displayName}`);
  console.log(`Metadata schema uri : ${result.metadataSchemaUri}`);
  console.log(`Metadata : ${JSON.stringify(result.metadata)}`);
  console.log(`Labels : ${JSON.stringify(result.labels)}`);
}
createDatasetVideo();

Python

To learn how to install or update the Vertex AI SDK for Python, see Install the Vertex AI SDK for Python. For more information, see the Python API reference documentation.

The following sample uses the Vertex AI SDK for Python to both create a dataset and import data. If you run this sample code, then you can skip the Import data section of this guide.

This particular sample imports data for classification. If your model has a different objective, then you must adjust the code.

def create_and_import_dataset_video_sample(
    project: str,
    location: str,
    display_name: str,
    src_uris: Union[str, List[str]],
    sync: bool = True,
):
    aiplatform.init(project=project, location=location)

    ds = aiplatform.VideoDataset.create(
        display_name=display_name,
        gcs_source=src_uris,
        import_schema_uri=aiplatform.schema.dataset.ioformat.video.classification,
        sync=sync,
    )

    ds.wait()

    print(ds.display_name)
    print(ds.resource_name)
    return ds

Import data

After you create an empty dataset you can import your data into the dataset. If you used the Vertex AI SDK for Python to create the dataset, then you might have already imported data when you created the dataset. If so, you can skip this section.

REST

Before using any of the request data, make the following replacements:

  • LOCATION: Region where dataset is to be stored. For example, us-central1.
  • PROJECT: Your project ID.
  • DATASET_ID: ID of the dataset.
  • IMPORT_FILE_URI: Path to the CSV or JSON Lines file in Cloud Storage that lists data items stored in Cloud Storage to use for model training; for import file formats and limitations, see Preparing video data.
  • OBJECTIVE: Designate either "classification", "object_tracking" or "action recognition" model objective.
  • PROJECT_NUMBER: Your project's automatically generated project number.

HTTP method and URL:

POST https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT/locations/LOCATION/datasets/DATASET_ID:import

Request JSON body:

{
  "import_configs": [
    {
      "gcs_source": {
        "uris": "IMPORT_FILE_URI"
      },
     "import_schema_uri" : "gs://google-cloud-aiplatform/schema/dataset/ioformat/automl_video_OBJECTIVE_io_format_1.0.0.yaml"
    }
  ]
}

To send your request, choose one of these options:

curl

Save the request body in a file named request.json, and execute the following command:

curl -X POST \
-H "Authorization: Bearer $(gcloud auth print-access-token)" \
-H "Content-Type: application/json; charset=utf-8" \
-d @request.json \
"https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT/locations/LOCATION/datasets/DATASET_ID:import"

PowerShell

Save the request body in a file named request.json, and execute the following command:

$cred = gcloud auth print-access-token
$headers = @{ "Authorization" = "Bearer $cred" }

Invoke-WebRequest `
-Method POST `
-Headers $headers `
-ContentType: "application/json; charset=utf-8" `
-InFile request.json `
-Uri "https://LOCATION-aiplatform.googleapis.com/v1/projects/PROJECT/locations/LOCATION/datasets/DATASET_ID:import" | Select-Object -Expand Content

You should see output similar to the following. You can use the OPERATION_ID in the response to get the status of the operation.

{
  "name": "projects/PROJECT_NUMBER/locations/LOCATION/datasets/DATASET_ID/operations/OPERATION_ID",
  "metadata": {
    "@type": "type.googleapis.com/google.cloud.aiplatform.v1.ImportDataOperationMetadata",
    "genericMetadata": {
      "createTime": "2020-10-08T20:32:02.543801Z",
      "updateTime": "2020-10-08T20:32:02.543801Z"
    }
  }
}

Java

Before trying this sample, follow the Java setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Java API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.


import com.google.api.gax.longrunning.OperationFuture;
import com.google.cloud.aiplatform.v1.DatasetName;
import com.google.cloud.aiplatform.v1.DatasetServiceClient;
import com.google.cloud.aiplatform.v1.DatasetServiceSettings;
import com.google.cloud.aiplatform.v1.GcsSource;
import com.google.cloud.aiplatform.v1.ImportDataConfig;
import com.google.cloud.aiplatform.v1.ImportDataOperationMetadata;
import com.google.cloud.aiplatform.v1.ImportDataResponse;
import java.io.IOException;
import java.util.Collections;
import java.util.List;
import java.util.concurrent.ExecutionException;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.TimeoutException;

public class ImportDataVideoObjectTrackingSample {

  public static void main(String[] args)
      throws IOException, ExecutionException, InterruptedException, TimeoutException {
    String gcsSourceUri =
        "gs://YOUR_GCS_SOURCE_BUCKET/path_to_your_video_source/[file.csv/file.jsonl]";
    String project = "YOUR_PROJECT_ID";
    String datasetId = "YOUR_DATASET_ID";
    importDataVideObjectTracking(gcsSourceUri, project, datasetId);
  }

  static void importDataVideObjectTracking(String gcsSourceUri, String project, String datasetId)
      throws IOException, ExecutionException, InterruptedException, TimeoutException {
    DatasetServiceSettings datasetServiceSettings =
        DatasetServiceSettings.newBuilder()
            .setEndpoint("us-central1-aiplatform.googleapis.com:443")
            .build();

    // 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 (DatasetServiceClient datasetServiceClient =
        DatasetServiceClient.create(datasetServiceSettings)) {
      String location = "us-central1";
      String importSchemaUri =
          "gs://google-cloud-aiplatform/schema/dataset/ioformat/"
              + "video_object_tracking_io_format_1.0.0.yaml";

      GcsSource.Builder gcsSource = GcsSource.newBuilder();
      gcsSource.addUris(gcsSourceUri);
      DatasetName datasetName = DatasetName.of(project, location, datasetId);
      List<ImportDataConfig> importDataConfigs =
          Collections.singletonList(
              ImportDataConfig.newBuilder()
                  .setGcsSource(gcsSource)
                  .setImportSchemaUri(importSchemaUri)
                  .build());

      OperationFuture<ImportDataResponse, ImportDataOperationMetadata> importDataResponseFuture =
          datasetServiceClient.importDataAsync(datasetName, importDataConfigs);
      System.out.format(
          "Operation name: %s\n", importDataResponseFuture.getInitialFuture().get().getName());
      System.out.println("Waiting for operation to finish...");
      ImportDataResponse importDataResponse = importDataResponseFuture.get(300, TimeUnit.SECONDS);

      System.out.format(
          "Import Data Video Object Tracking Response: %s\n", importDataResponse.toString());
    }
  }
}

Node.js

Before trying this sample, follow the Node.js setup instructions in the Vertex AI quickstart using client libraries. For more information, see the Vertex AI Node.js API reference documentation.

To authenticate to Vertex AI, set up Application Default Credentials. For more information, see Set up authentication for a local development environment.

/**
 * TODO(developer): Uncomment these variables before running the sample.\
 */

// const datasetId = 'YOUR_DATASET_ID';
// const gcsSourceUri = 'YOUR_GCS_SOURCE_URI';
// eg. 'gs://<your-gcs-bucket>/<import_source_path>/[file.csv/file.jsonl]'
// const project = 'YOUR_PROJECT_ID';
// const location = 'YOUR_PROJECT_LOCATION';

// Imports the Google Cloud Dataset Service Client library
const {DatasetServiceClient} = require('@google-cloud/aiplatform');

// Specifies the location of the api endpoint
const clientOptions = {
  apiEndpoint: 'us-central1-aiplatform.googleapis.com',
};
const datasetServiceClient = new DatasetServiceClient(clientOptions);

async function importDataVideoObjectTracking() {
  const name = datasetServiceClient.datasetPath(project, location, datasetId);
  // Here we use only one import config with one source
  const importConfigs = [
    {
      gcsSource: {uris: [gcsSourceUri]},
      importSchemaUri:
        'gs://google-cloud-aiplatform/schema/dataset/ioformat/video_object_tracking_io_format_1.0.0.yaml',
    },
  ];
  const request = {
    name,
    importConfigs,
  };

  // Create Import Data Request
  const [response] = await datasetServiceClient.importData(request);
  console.log(`Long running operation: ${JSON.stringify(response.name)}`);

  // Wait for operation to complete
  const [importDataResponse] = await response.promise();

  console.log(
    `Import data video object tracking response : \
      ${JSON.stringify(importDataResponse)}`
  );
}
importDataVideoObjectTracking();

Python

To learn how to install or update the Vertex AI SDK for Python, see Install the Vertex AI SDK for Python. For more information, see the Python API reference documentation.

def import_data_video_object_tracking_sample(
    project: str,
    location: str,
    dataset_name: str,
    src_uris: Union[str, List[str]],
    sync: bool = True,
):
    aiplatform.init(project=project, location=location)

    ds = aiplatform.VideoDataset(dataset_name=dataset_name)

    ds.import_data(
        gcs_source=src_uris,
        import_schema_uri=aiplatform.schema.dataset.ioformat.video.object_tracking,
        sync=sync,
    )

    ds.wait()

    print(ds.display_name)
    print(ds.resource_name)
    return ds

Get operation status

Some requests start long-running operations that require time to complete. These requests return an operation name, which you can use to view the operation's status or cancel the operation. Vertex AI provides helper methods to make calls against long-running operations. For more information, see Working with long-running operations.