Skip to main content

Understanding Automatic Redaction via VIDIZMO Vision Indexer

Overview​

VIDIZMO Vision Indexer enables users to perform automatic object detection on their Portal content. It employs cutting-edge artificial intelligence (AI) models to identify objects such as faces, vehicles, devices, and more.

One of the key features of the Vision Indexer is automatic redaction, which allows users to hide sensitive content within their Portal content without manual intervention. This feature is particularly useful for organizations that must comply with privacy regulations, such as GDPR, or wish to protect sensitive information from being exposed.

Automatic redaction using VIDIZMO Vision Indexer streamlines the content review process and ensures consistent, high-quality redactions across vast content. By automating this process, organizations can reduce the time spent manually reviewing media while ensuring that sensitive data is appropriately handled.

Concept​

Redaction of Visual Insights via Automatic Processing​

Redaction of visual insights refers to the process of automatically detecting and removing specified objects from your Portal content. When automatic processing is enabled on the VIDIZMO Vision Indexer, the content you upload to the Portal is automatically processed according to its configuration.

The Vision Indexer app first performs the detection of objects that are specified in its configurations. Once the detection is complete, the application automatically applies redaction only on the objects selected for redaction in its configurations. This process ensures that private or sensitive data is not exposed while maintaining the content's integrity for other purposes, such as content analysis, distribution, or archiving.

If you want to see the steps involving automatic processing, see Automatic Redaction of Visual Insights.

Redaction of Visual Insights via On-Demand Detection Processing​

In addition to automatic redaction, VIDIZMO supports on-demand detection and redaction, offering users greater flexibility and control over processing their Portal content. With on-demand processing, users can selectively perform detection and redaction on content they upload via custom upload or existing content (individually or in bulk) via the process modal.

This selective approach means the preset configurations of the Vision Indexer do not restrict users. Instead, they can specify which objects to detect and redact at any time based on their immediate needs. This is particularly useful when redacting sensitive information in specific videos or documents without applying broad automatic settings across all content.

Moreover, on-demand redaction eliminates the need to manually enter Studio Space and redact objects step-by-step. Users can efficiently redact specific objects from their Portal content directly through the on-demand workflows, saving significant time and effort.

If you want to see the steps involved in on-demand processing, see Automatic Redaction of Visual Insights.

Workflow for the Redaction Process​

The redaction process in VIDIZMO Vision Indexer is designed to be straightforward and efficient. It follows a series of sequential steps, from object detection to the final application of redaction.

A workflow is initiated whenever you process your content for detection or redaction. In VIDIZMO, workflows represent activities executed in a defined sequence to complete specific operations. These workflows automate complex processes such as media indexing, transcoding, or redaction, ensuring consistency and reducing manual effort.

For more detailed information on workflows, see Understanding Workflows

When a redaction workflow is triggered, the system performs the following key activities:

  • Detection Activity (Indexing): The Vision Indexer scans the media content to detect objects based on user-defined categories (e.g., faces, license plates).

  • VIDIZMO Redaction: Once the detection is complete, the system applies redaction on the detected objects according to the user's selection.

You can monitor the status and progress of redaction workflows via the Control Panel > Workflows page, where each workflow's current state, completion percentage, and errors can be viewed.

Redaction Threshold​

The redaction threshold setting is essential in determining which detected objects are automatically redacted by the VIDIZMO Vision Indexer. It allows users to control the sensitivity of the redaction process for different detection types.

It must be set to a value equal to or higher than the detection threshold. Only objects with confidence scores at or above this level are redacted during the redaction process.

Detections with confidence scores below the redaction threshold remain untouched. Users can manually review and redact them if desired, allowing them to verify lower-confidence objects selectively. This ensures that no sensitive content is overlooked due to strict automatic settings.

A higher threshold reduces false positives (incorrect redactions) but may miss some sensitive objects, especially if the content quality is low or objects are partially obscured.

A lower threshold increases detection sensitivity but risks redacting non-sensitive or irrelevant content due to false positives.

Balancing the confidence threshold is essential to maintaining redaction accuracy and preventing unnecessary loss of information.

Automatic Redaction of Documents​

VIDIZMO Vision Indexer allows you to automatically redact specific information in your documents with the help of custom patterns. This functionality is only supported for editable documents (the ones with selectable text).

Custom patterns can be created on your Portal using plain-text strings or regular expressions. VIDIZMO also provides you with some custom patterns by default.

The automatic document process involves scanning documents on your Portal against the custom patterns selected in the Vision Indexer. These custom patterns enable precise targeting and redaction of specific data like social security numbers, credit card numbers, email addresses, or other sensitive terms relevant to your use case.

Advanced Processing Options​

Advanced processing options allow users to control how the system handles the file during the automatic redaction process. If these advanced processing options are configured in the application, they will be applied in both automatic processing during upload and on-demand processing. As of now, these options include:

Action for Original File: Decides whether to retain, delete, or override the original file after redaction.

  • Retain File​

When this option is selected, a copy of the original content is created specifically for redaction purposes. The original (main) file retains all the detected objects per the detection settings, while the redacted copy only excludes the objects chosen for redaction.

The processing workflow includes the following:

  1. The main file undergoes detection for the specified object types.
  2. A duplicate copy of the main file is created.
  3. Metadata containing all detections is copied over to the duplicate.
  4. Redaction only applies to the duplicate copy, removing the selected objects.
  5. The original file remains unaltered, with all detections intact.

This approach allows organizations to maintain an unaltered original version for purposes (such as compliance or auditing) while still providing a redacted version suitable for distribution or other use cases.

  • Delete and Move to Recycle Bin​

When this option is selected, a duplicate of the original file is created specifically for redaction. Once the redaction process is complete, the original file is deleted and moved to the Recycle Bin. This option ensures that only the redacted version of the file remains published and available in your Portal library.

The processing workflow includes the following:

  1. The original file undergoes detection for the specified object types.
  2. A duplicate copy of the main file is created for redaction.
  3. Metadata containing all detections is copied over to the duplicate.
  4. Redaction is applied to the duplicate copy, removing the selected objects.
  5. The original file is deleted and sent to the Recycle Bin

  • Override Original File​

When this option is selected, the detection and redaction process is performed directly on the original file, modifying it in place. No duplicate is created, so the original file is altered as part of the redaction process.

The processing workflow includes the following:

  1. The original file undergoes detection for the specified object types.
  2. Redaction is applied directly to the original file, removing the selected objects.
  3. The original file is altered, with metadata or timed data updated part of the process.

If you want to know how these settings aare configured, refer to the Advanced Processing section in Configuring VIDIZMO Vision Indexer for Object Detection

Applying Advanced Processing Options​

The advanced processing options are only applied when automatic processing is enabled in the VIDIZMO Vision Indexer, regardless of whether the user performs automatic or on-demand processing.

  • When Automatic Processing is Enabled, the content undergoing processing for automatic redaction is handled according to the selected advanced processing option.  
  • However, if Automatic Processing is Disabled, the automatic redaction process will revert to the default behavior for handling original files, which is to create a redacted copy, similar to the Retain File option.
Did you find this article helpful?