Authoring Information
The Annotation Authoring Information feature provides comprehensive tracking and visibility into the creation and modification history of annotations (cuboids, 3D polylines, and 2D bounding boxes). This feature captures critical metadata about who created or modified an annotation, when changes were made, how the annotation was generated, and where in the workflow it was edited.
Key information tracked includes:
Timestamp: Creation and edit dates/times for each annotation
User Identity: Email ID or employee ID of the creator/editor
Annotation Source: Origin method (prelabel, manual drawing, smart automation like interpolation or dynamic sizing)
Workflow Stage: The specific stage/node where changes occurred
Change History: All attribute modifications, including positional, dimensional, rotational, and property changes
Visual Comparison: Side-by-side visualization of any two historical versions
The system consolidates changes per session (stage + labeler combination) to show the final state rather than every incremental edit, providing a clear audit trail without overwhelming detail.
Track who changed what, when, and how—then visually compare any two versions to validate annotation quality instantly.
Use Cases
Quality Assurance & Review: Quality reviewers can trace annotation lineage to identify patterns in errors, verify that annotations meet quality standards, and understand which automated tools or manual processes produced specific results.
Dispute Resolution: When questions arise about annotation accuracy or consistency, teams can review the complete history to understand decision-making context and resolve disagreements with factual evidence.
Debugging & Issue Investigation: When annotation quality issues are detected downstream, teams can trace back through the history to identify where problems were introduced and prevent recurrence.
Benefits
Good: Reviewers can quickly identify suspicious changes, verify automation accuracy, and ensure annotations meet project standards, resulting in higher-quality datasets.
Fast: Reduce time spent investigating quality issues from hours to minutes by quickly pinpointing when and where problems occurred.
Cost Efficient:
Identify and address quality issues early before they propagate through the pipeline, avoiding expensive late-stage corrections.
Analyze prelabel vs. manual annotation performance to make data-driven decisions about where automation delivers ROI and where human expertise is essential.
Understand which stages and annotators require more support, allowing managers to allocate training, supervision, and resources where they'll have the greatest impact.
How to Use the Authoring Information
Accessing Authoring Information
Click on any cuboid, 3D polyline, or 2D bounding box
Right-click and navigate to the Authoring Information. The Authoring Information panel appears
The authoring information panel displays the following information sorted on the recent changes:
Creation timestamp and creator identity
Most recent edit timestamp and editor identity
Annotation source (prelabel/manual/smart automation)
Workflow stage where changes occurred
List of all attribute changes made
Understanding the Information Display
User Identity: Shows email address when available, otherwise displays employee ID for attribution.
Source Indicators:
Prelabel: Annotation originated from automated prelabeling pipeline
Manual: Created by human labeler from scratch
Automation: Generated or modified by tools like interpolation, dynamic sizing, or propagation
Stage Information: Displays the workflow node (e.g., "Labeling/OP", "Review/QC") where the annotation reached its current state in that session.
Change Consolidation: Only the final state per session (stage + labeler) is shown, not every intermediate adjustment.
Visual Comparison Feature
In the authoring information panel, click on the toggle beside each version
The system displays both versions simultaneously with the version history appearing on the right panel
Switch between split-screen, overlay, and difference highlighting modes
Best Practices
For Project Managers
Regular Audits: Schedule periodic reviews of authoring information to identify process bottlenecks, training needs, or quality issues before they escalate.
Leverage Source Data: Analyze prelabel vs. manual annotation metrics to optimize automation investment and understand where human expertise adds most value.
Set Expectations: Communicate to teams that authoring information is tracked for quality improvement, not punitive measures, to maintain a positive learning culture.
For Quality Reviewers
Start with Source: When reviewing annotations, check the source first—prelabeled items may need closer scrutiny than manually created annotations that have been through multiple stages.
Compare Versions Strategically: Use visual comparison between initial creation and current state to quickly spot significant changes that warrant investigation.
Document Patterns: When you notice recurring issues from specific sources, stages, or individuals, document these patterns to drive systematic improvements.
Validate Automation: Regularly sample smart automation results to ensure interpolation, dynamic sizing, and other tools maintain accuracy.
General Best Practices
Version Comparison for Validation: Always use visual comparison when investigating annotation discrepancies.
Session Consolidation Awareness: Remember that only final session states are shown—if you need more granular detail, review changes in real-time rather than retroactively.
Document Exceptional Cases: When unusual patterns appear in authoring history (e.g., multiple rapid edits), investigate and document root causes to prevent future issues.
Last updated