Label data using the Google Cloud console

For text, image, and video data, you can import labeled or unlabeled data and add labels using the Google Cloud console. You can also delete or add new labels to existing labeled datasets.

To learn how to import your data, see the Prepare data page of the data type and objective that you're working with on the Training overview page. Continue with the respective Create dataset page for your data type and objective.

After creating the dataset and importing the unlabeled data, you will be in Browse mode.

How to add labels

Instructions for labeling objectives are provided here for each data type.

Image

The images you have just imported in the dataset are unlabeled, as expected.

Classification

When in Browse mode, and the dataset with the unlabeled images is selected, you can see your uploaded images.

1. Click Add new label and enter your new label.
2. Click Done.
   Repeat for each label you want to add.
3. Select the image you want to label.
   The list of labels appears.
4. Select the label you want to associate with the image.
5. Click Save.

Classification

When in Browse mode, and the dataset with the unlabeled images is selected, you can see your uploaded images.

1. Click Add new label and enter your new label.
2. Click Done.
   Repeat for each label you want to add.
3. Select the image you want to label.
   The list of labels appears.
4. Select the label you want to associate with the image.
5. Click Save.
6. You can view the labels applied to each image in the Browse tab.

Object detection

When in Browse mode, and the dataset with the unlabeled images is selected, you can see your uploaded images.

1. Click Add new label and enter your new label.
2. Click Done.
   Repeat for each label you want to add.
3. Select the image you want to label.
4. The list of labels objects appears, if there are any.
5. In the add annotation window, select the Add bounding box button to add an object bounding box to the image.
   
6. After drawing a bounding box, a list of labels to apply to the object will appear. Choose the appropriate label.
   
7. After you have added all labels and bounding boxes, click Save to update the image's annotations.
   

Text

Starting on September 15, 2024, you can only customize classification, entity extraction, and sentiment analysis objectives by moving to Vertex AI Gemini prompts and tuning. Training or updating models for Vertex AI AutoML for Text classification, entity extraction, and sentiment analysis objectives will no longer be available. You can continue using existing Vertex AI AutoML Text models until June 15, 2025. For a comparison of AutoML text and Gemini, see Gemini for AutoML text users. For more information about how Gemini offers enhanced user experience through improved prompting capabilities, see Introduction to tuning. To get started with tuning, see Model tuning for Gemini text models

The text you have just imported in the dataset may or may not be labeled.

Single-label classification

1. Click the text to label.
2. To add a new label, click Add label.
3. For a single-label classification dataset, select one label to associate with the text.
   You can delete labels, as well.
4. Click Save.

Multi-label classification

1. Click the text to label.
2. To add a new label, click Add label.
3. For multi-label classification datasets, select one or more labels to associate with the text.
   You can delete labels, as well.
4. Click Save.

Entity extraction

1. Click the text to label.
2. To add a new label, click Add label.
3. Select one or more words to label.
4. Select a label to associate with the selected text.
5. Click Save.

Sentiment analysis

1. Click the text to label.
2. To add a new label, click Add label.
3. Select a score to associate with the text.
4. Click Save.

Video

The videos you have imported in the dataset are unlabeled, as expected.

1. To navigate to your new dataset, click "Datasets" in the navigation menu.
   
2. Select the dataset you want to add labels to.
   Dataset appears.
   

Action recognition

When in Browse mode, and the dataset with the unlabeled videos is selected, you should see your videos.

1. Add label(s).
   
2. Select a video and start viewing.
3. When the action starts appearing that you want to identify, slowly progress through till you find the center or the most representative moment of the action using "Next frame" option.
4. Click Add annotation.
   Your list of labels appears.
5. Select the label you want for this video segment.
6. Click Save.

Classification

When the dataset with the unlabeled videos is selected, you will see your videos.

1. Add labels.
   
2. Select the video you want to label. The video appears with your list of color-coded labels below.
3. Navigate to starting point of segment. Click Add segment.
   1. Select segment. Adjust to designate the segment to be used for training for your particular label:
      
   2. Select label. You can select multiple labels for a segment.
      
   3. Click Done, then Save.
   4. Repeat steps to add another label to the same video with a different time segment. Here's an example of where two more labels were added:
      
4. Return to Datasets list to repeat.

Object tracking

When in BROWSE mode, and the dataset with the unlabeled videos is selected, you should see your videos.

1. Click Add Label and add the labels you plan to use (for example, "sedan", "pickup", "SUV").
2. Click Save. If you need to add more labels later, click Add new label and Save.
3. Select a video and start viewing.
4. When an object appears that you want to track,
   1. Stop the video.
   2. Drag a bounding box from the upper left corner, down to the lower right corner. Be sure to make the box as tight as possible around the object.
      A list of labels appears to the lower right-hand side of the bounding box.
   3. Select the appropriate label.
      Note: Note: You can always add more labels during this process. Be sure to save the added label so it shows up in the list going forward.
   4. Click Save.

What's next

• Training your AutoML model using the Google Cloud console.

• Training your AutoML Edge model using the Google Cloud console. (image and video only)

• Training your AutoML model using the Vertex AI API.

• Training your AutoML Edge model using the Vertex AI API. (image and video only)