MultiCam Grid View

The MultiCam Grid View is designed for efficient cross-camera, multi-frame object review and validation. After annotators complete 2D annotations and tracking across multiple camera sensors for a single object, this feature provides a consolidated view to verify annotation quality and consistency.

This feature is particularly valuable when:

  • Validating object tracking across multiple synchronized camera feeds

  • Reviewing annotation consistency for the same object captured from different angles

  • Performing quality assurance on completed annotation tasks

  • Identifying gaps or errors in multi-camera object tracking


Benefits

  • Fast: View all annotations related to a single object across multiple camera sensors and frame sequences in one unified interface, eliminating time-consuming context switching between individual camera views.

  • Good: Improve annotation quality and consistency through comprehensive visual validation. The intuitive grid layout organizes data logically—rows represent sequential frames while columns represent different camera sensors—making it easy to spot inconsistencies, gaps, or errors at a glance.

  • Cost Effective: Lower operational costs by minimizing rework and reducing quality assurance time.


Steps to Use MultiCamGrid View

Accessing MultiCam Grid View

  1. Create your annotation by tracking 2D objects as usual using your standard annotation workflow.

  2. Select the object you wish to review and click on MultiCamGrid View on the Main Menu

  3. Click on any annotation within the cell, it will activate the Annotation Details interface

    • Modify attribute values through the Annotation Details pop-up

    • Hit Save to save attribute changes

  4. Resize the bounding box or 2D polygon by dragging the corners or edges

  5. Translate the annotation by clicking and dragging

  6. Changes are saved automatically

Settings

  • Rows: Each row represents a frame in the sequence

  • Columns: Each column represents a different camera sensor (refer to the column name and #-ing index)

  • Empty Cells: Display "no object" message when the object is not present in that camera-frame combination

Customizing the Display

  1. Adjust the number of camera sensor columns visible on screen using the layout settings (Images per row slider)

  2. Configure based on your monitor size for optimal viewing experience

Reviewing Attributes

  1. Enable attribute display from the view options

  2. Select specific attributes to display, or choose to show all attributes

  3. Colored tags will appear at the bottom edge of each image showing the attribute values

  4. Compare attribute consistency across frames and cameras at a glance


Best Practices

Before Using MultiCam Grid View

  • Ensure 2D annotation and tracking is finished across all relevant camera sensors before entering grid view

  • Work with one object at a time for focused review and validation

During Review

  • Review frame by frame, checking each camera sensor's annotation for consistency

  • Use the attribute display feature to quickly identify inconsistencies in classification or properties across views

  • Pay attention to "no object" indicators, verify whether these represent true absence or missing annotations

Optimizing Display

  • Start with a column count that fits comfortably on your monitor without requiring horizontal scrolling and also enables you to see the annotation. Find that balance.

  • When reviewing specific properties, display only relevant attributes to reduce visual clutter

  • For large sequences, break the review into manageable chunks rather than attempting to validate everything at once

Quality Assurance

  • Check that the same object shows consistent positioning relative to its environment across different camera angles

  • Verify smooth transitions in bounding box position and size across sequential frames

  • Ensure attribute values remain consistent unless there's a valid reason for changes (e.g., occlusion state)

Editing Guidelines

  • Use click-to-edit for fine-tuning rather than major corrections—significant errors may require returning to the standard annotation view

Last updated