Updated on 2024-04-30 GMT+08:00

Object Detection

Training a model uses a large number of labeled images. Therefore, label images before the model training. You can add labels to images by manual labeling or auto labeling. In addition, you can modify the labels of images, or remove their labels and label the images again.

Before labeling an image in object detection scenarios, pay attention to the following:

  • All target objects in the image must be labeled.
  • Target objects are clear without any blocking and contained within bounding boxes.
  • Only the entire object must be contained within a bounding box. The bounding box contains the entire object. The edge of the bounding box cannot intersect the edge outline of the object to be labeled. Ensure that there is no gap between the edge and the object to be labeled to prevent the background from interfering with the model training.

Starting Labeling

  1. Log in to the ModelArts management console. In the navigation pane on the left, choose Data Management > Label Data.

    Data management is being upgraded and is invisible to users who have not used data management. To use data management, you are advised to submit a service ticket to obtain the permissions.

  2. In the labeling job list, select a labeling type from the All type drop-down list, click the job to be performed based on the labeling type. The details page of the job is displayed.
    Figure 1 Selecting a labeling type
  3. The job details page displays all data of the labeling job.

Synchronizing New Data

ModelArts automatically synchronizes data and labeling information from datasets to the labeling job.

To quickly obtain the latest data in a dataset, on the All statuses, Unlabeled, or Labeled tab page of the labeling job details page, click Synchronize New Data.

Symptom:

After the labeled data is uploaded to OBS and synchronized, the data is displayed as unlabeled.

Possible causes:

Automatic encryption is enabled in the OBS bucket.

Solution:

Create an OBS bucket and upload data again, or disable bucket encryption and upload data again.

Filtering Data

On the All statuses, Unlabeled, or tab page, click in the filter criteria area and add filter criteria to quickly filter the data you want to view.

The following filter criteria are available. You can set one or more filter criteria.

  • Example Type: Select Hard example or Non-hard example.
  • Label: Select All or one or more labels you specified.
  • File Name or Path: Filter files by file name or file storage path.
  • Labeled By: Select the name of the user who labeled the image.
  • Sample Attribute: Select the attribute generated by auto grouping. This filter criterion can be used only after auto grouping is enabled.
  • Data Attribute: Select All or Inference to filter the data source.
Figure 2 Filter criteria

Manually Labeling Images

The labeling job details page displays the All statuses, Unlabeled, and Labeled tab pages. The Unlabeled tab page is displayed by default.

  1. On the Unlabeled tab page, click an image. The system automatically directs you to the page for labeling the image. For details about how to use common buttons on this page, see Table 2.
  2. In the tool bar, select a proper labeling shape. The default labeling shape is a rectangle. In this example, the rectangle is used for labeling.

    In the tool bar, multiple tools are provided for you to label images. After you select a shape to label the first image, the shape automatically applies to subsequent images. You can switch the shape as required.

    Table 1 Supported bounding box

    Icon

    Description

    Rectangle. You can also press 1. Click the edge of the upper left corner of the object to be labeled. A rectangle will be displayed. Drag the rectangle to cover the object and click to label the object.

    Polygon. You can also press 2. In the area where the object to be labeled is located, click to label a point, move the mouse and click multiple points along the edge of the object, and then click the first point again. All the points form a polygon. In this way, the object to be labeled is within the bounding box.

    Round. You can also press 3. Click the center point of an object, and move the mouse to draw a circle to cover the object and click to label the object.

    Straight. You can also press 4. Click to specify the start and end points of an object, and move the mouse to draw a straight line to cover the object and click to label the object.

    Dashed line. You can also press 5. Click to specify the start and end points of an object, and move the mouse to draw a dashed line to cover the object and click to label the object.

    Dot. You can also press 6. Click the object in an image to label a point.

  3. In the Add Label text box, enter a new label name, select the label color, and click Add. Alternatively, select an existing label from the drop-down list.
    Label all objects in an image. Multiple labels can be added to an image. After labeling an image, click the right arrow (or press D) in the upper right corner of the image to switch to the next image and label the image.
    Figure 3 Adding an object detection label
  4. Click Back to Data Labeling Preview in the upper left part of the page to view the labeling information. In the dialog box that is displayed, click Yes to save the labeling settings.

    The selected images are automatically moved to the Labeled tab page. On the Unlabeled and All statuses tab pages, the labeling information is updated along with the labeling process, including the added label names and the number of images for each label.

Table 2 Common icons on the labeling page

Button

Features

Cancel the previous operation. You can also press Ctrl+Z.

Redo the previous operation. You can also press Ctrl+Shift+Z.

Zoom in an image. You can also use the mouse wheel to zoom in.

Zoom out an image. You can also use the mouse wheel to zoom out.

Delete all bounding boxes on the current image. You can also press Shift+Delete.

Show or hide a bounding box. This operation can be performed only on a labeled image. You can also press Shift+H.

Drag a bounding box to another position or drag the edge of the bounding box to resize it. You can also use X + left mouse button.

Reset a bounding box. After dragging a bounding box, you can click this button to quickly restore the bounding box to its original shape and position. You can also press Esc.

Viewing Labeled Images

On the labeling job details page, click the Labeled tab to view the list of labeled images. The labels of each image are displayed below the image.

Figure 4 Labels

Quick Review

To simplify operations, ModelArts provides quick review so that you can batch review and modify labeled data.

  1. Log in to the ModelArts management console. In the navigation pane, choose Data Management > Label Data. On the My Creations tab page, select the target labeling job type from the All types drop-down list in the upper right corner. (Only object detection and image segmentation support quick review.)
  2. In the labeling job list, click the target labeling job. The labeling details page is displayed.
  3. Click Quick Review on the Labeled tab. On the displayed page, confirm the labeling results.
    Figure 5 Quick Review
  4. Batch review images of the same label.
    1. On the review page, select the label type from the drop-down list next to Filter by Label.
    2. Sort images of the selected label type by bounding box area or aspect ratio.
    3. Click an incorrectly labeled image, and then drag the labeling box to relabel the image. (Modified is displayed on the modified images.)
    4. You can select the incorrectly labeled images, and then click in the upper right corner to delete the label. (Deleted is displayed on the images whose label has been deleted.)
      Figure 6 Modified
      Figure 7 Deleted
    5. You can also modify the label of a labeled image.
      1. Select the target images and click in the All Labels area on the right.
      2. Type a new label and click OK.
        Figure 8 All Labels
        Figure 9 Adding a label
  5. After the modification, click Apply Modifications. In the displayed dialog box, click OK. The system automatically returns to the labeling overview page and overwrites the original labeling data.
    Figure 10 Apply Modifications
  6. If you are not satisfied with the modified data, you can click Cancel Modifications to retain the original labeling data.
    Figure 11 Cancel Modifications
    Table 3 Buttons on the quick review page

    Button

    Features

    Delete the label.

    Undo all operations on the current page.

    Undo the previous operation.

    Redo the previous operation.

Modifying Labeled Data

After labeling data, you can modify labeled data on the Labeled tab page.

  • Modifying based on images

    On the labeling job details page, click the Labeled tab and then the image to be modified. The labeling page is displayed. Modify the image information in the label information area on the right.

    • Modifying a label: In the Labeling area, click the edit icon, enter the correct label name in the text box, and click the check mark to complete the modification. Alternatively, click a label. In the image labeling area, adjust the position and size of the labeling box. After the adjustment, right-click the labeling box and choose Modify from the shortcut menu. Enter the new label and click Modify to save the modification.
    • Deleting a label: In the Labeling area, click the deletion icon to delete a label from the image.

      After deleting the label, click Back to Data Labeling Preview in the upper left corner of the page to exit the labeling page. In the dialog box that is displayed, save the modification. After all labels of an image are deleted, the image is displayed on the Unlabeled tab page.

      Figure 12 Editing an object detection label
  • Modifying based on labels
    • On the labeling job details page, click Label Management on the right. All label information is displayed.
      • Modifying a label: Click Modify in the Operation column. In the dialog box that is displayed, enter a new label name, select a new label color, and click OK. After the modification, the images with the label added will use the new label name.
      • Deleting a label: Click Delete in the Operation column, or select the label to be deleted and click Delete Label above the label list.
      Figure 13 Label Management
      Figure 14 All labels
    • Alternatively, click Label in the Operation column of the target labeling job to go to the label management page.
      Figure 15 Accessing the label management page from the labeling job list
      • Click Modify in the Operation column of the target label to modify it.
      • Click Delete in the Operation column of the target label to delete it.

Adding Data

In addition to the data automatically synchronized from datasets, you can directly add images to labeling jobs for labeling. The added data is first imported to the dataset associated with the labeling job. Then, the labeling job automatically synchronizes the latest data from the dataset.

  1. On the labeling job details page, click All statuses, Labeled, or Unlabeled tab, click Add data in the upper left corner.
    Figure 16 Adding data
  2. Configure the data source, import mode, import path, and labeling status.
    Figure 17 Adding images

  3. Click OK.

    The images you have added will be automatically displayed in the image list on the All statuses tab page. You can choose Add data > View historical records to view task history.

    Figure 18 Viewing historical data

Deleting Images

You can quickly delete the images you want to discard.

On the All statuses, Unlabeled, or Labeled tab page, select the images to be deleted or click Select Images on Current Page, and click Delete. In the displayed dialog box, select or deselect Delete the source files from OBS as required. After confirmation, click Yes to delete the images.

Figure 19 Deleting images

If a tick is displayed in the upper left corner of an image, the image is selected. If no image is selected on the page, the Delete button is unavailable.

If you select Delete the source files from OBS, images stored in the OBS directory will be deleted accordingly. This operation may affect other dataset versions or datasets using those files, for example, leading to an error in page display, training, or inference. Deleted data cannot be recovered. Exercise caution when performing this operation.

Managing Annotators

If team labeling is enabled for a labeling job, view its labeling details on the Annotator Management tab page. Additionally, you can add, modify, or delete annotators.

  1. Choose Data Management > Label Data. On the My Creations tab page, view the list of all labeling jobs.
  2. Locate the row that contains the target team labeling job. (The name of a team labeling job is followed by .)
  3. Choose More > Annotator Management in the Operation column. Alternatively, click the job name to go to the job details page, and choose Team Labeling > Annotator Management in the upper right corner.
    Figure 20 Annotator management (1)
    Figure 21 Annotator management (2)

  • Adding an annotator

    Click Add Member, select a member name, and click OK.

    Click Send Email in the Operation column to send the labeling job to the annotator by email.

  • Modifying annotator information

    Click Modify in the Operation column to modify the role of the annotator.

  • Deleting an annotator

    Click Delete in the Operation column to delete the annotator.