Skip to main content
Version: V12

Configuring VIDIZMO Vision Indexer for Object Detection

The VIDIZMO Vision Indexer uses AI to process videos, images, and documents, extracting insights such as detected objects and recognized characters. These insights improve content categorization, indexing, and search, making it easier to find relevant results in your Portal.

You can configure the Vision Indexer to automatically process AI insights and apply settings that match your workflow. Options include automatic redaction of selected insights and control over how processed output files are handled.

Prerequisites​

To configure the VIDIZMO Vision Indexer, ensure:

  • You belong to a group with feature permissions to configure the VIDIZMO Vision Indexer.
  • You have specific feature permissions for each detection type you intend to configure. For example, to configure Face Detection settings, your group must have permissions enabled for the face detection feature.

For details about managing groups and enabling feature permissions, see Enable Features in the VIDIZMO Portal.

  1. In VIDIZMO, select the menu icon in the top-left corner to open the navigation pane.
  2. Expand the Admin section and select Portal Settings.
  3. Go to Apps > Content Processing.
  4. Select the settings icon on the VIDIZMO Vision Indexer app.

Object Detection Configuration​

Media Formats​

  1. Media Formats Select the available media or evidence type. Choose from:

    • Image
    • Video
    • Document

    In a DEMS Portal, this field appears as Evidence Formats.

NOTE: VIDIZMO also supports images with equirectangular projection (such as panoramic images) for object detection.

The configuration screen dynamically shows fields based on the selected media formats and detection types.

Detection Types​

  1. Detection Types Select the required detection method depending on your use case:

    • Object Detection Detects visual objects in content.
    • OCR Optical Character Recognition for text extraction.

    You can select both detection types to use them together.

Object Detection Classes​

  1. Object Detection Classes Select the object classes you want to detect. The VIDIZMO Vision Indexer can detect the following:
Face
Person
Gun
Knife
Train
Car
Bus
Truck
Boat
Bike
Airplane
License Plate
Screen
Laptop
Display
Cellphone
Signature
Street Sign
Head
Eyes Protection
No Eyes Protection
Nose Mouth Protection
No Nose Mouth Protection
Head Protection
No Head Protection
Hand Protection
No Hand Protection
Foot Protection
No Foot Protection
Safety Vest
No Safety Vest
House Number
Pothole
Fallen Tree
Graffiti
Fire Extinguisher
Bulletproof Vest
Mailbox
Tattoo
Fire
Smoke
Traffic Light
Road Signs
Traffic Cone
Zebra Crossing

When you select one or more object detection classes, a configuration field appears for each selected class.

Confidence Threshold​

  1. Confidence Threshold Set the minimum confidence score required for an object to be detected. Higher values increase precision but may reduce the number of detected instances, while lower values increase sensitivity.

    Range: 10 to 90. Default: 45.

    You can configure a separate confidence threshold for each selected object detection class.

Face Detection Settings​

  1. Facial Attributes When Face is selected as a detection class, select the facial attributes you want to detect:
    • Age
    • Gender
    • Race

NOTE: The selected facial attributes are detected when a video or image undergoes Face Detection processing. After processing, the detected attributes are displayed on the processed file's playback page.

Video Processing Settings​

The following settings apply specifically to video content processing.

  1. Tracking Frame Specify the number of consecutive frames in which an object must be detected before it is considered valid. The model disregards objects that do not appear in the specified number of consecutive frames.

    Range: 3 to 30. Default: 8.

  2. Frame Forward Count Control the stride of video analysis for visual objects (for example, Street Sign, Face, Person). After VIDIZMO processes one video frame, it moves forward by this many frames before processing the next one.

    Range: 1 to 15. Default: 1.

    • Lower values analyze frames more frequently (higher accuracy, less redaction flicker).
    • Higher values skip more frames (faster processing, but can miss brief objects).

  3. OCR Frame Forward Count Control the stride for OCR (text or character) analysis in videos. After VIDIZMO completes OCR on one frame, it moves forward by this many frames before running OCR again.

    Range: 1 to 30. Default: 15.

    • Lower values capture transient or small text more reliably.
    • Higher values speed up OCR on long recordings but may miss fleeting text.

NOTE: This field appears only when OCR is selected as a detection type.

OCR Language​

  1. OCR Language Select the language for Optical Character Recognition during processing. Choose the language that matches your content for optimal text extraction accuracy.

    For more details on OCR processing, see How to Perform OCR using VIDIZMO Vision Indexer.

Redaction Settings​

The VIDIZMO Vision Indexer includes settings for automatically redacting detected objects within your content.

NOTE: Redaction settings are only available if you belong to a group with Redaction or Bulk Redaction features enabled.

The automatic redactions are applied to detections generated either through automatic processing of uploaded media or through on-demand processing. Only the objects you have explicitly enabled for detection are subject to redaction.

Automatic Object Redaction​

  1. Redaction Type(s) Select the detection types that you want the indexer to redact automatically during content processing.

  2. Redaction Sub Type(s) Select specific subtypes for redaction. The available options depend on the Object Detection Classes you configured. Only the selected classes are processed for redaction.

  3. Confidence Threshold for Redaction Set the minimum confidence threshold for automatic redaction. Objects detected with confidence levels equal to or greater than this threshold are automatically redacted.

    Range: 10 to 99. Default: 10.

  4. Visual Masking Style Select how objects are redacted in your content:

    • Solid Fully blacks out the detected object.
    • Pixelate Distorts the content by replacing the detected area with a grid of blocks.
    • Blur Obscures the detected area by applying a blurring effect.

Automatic Document Redaction​

Document redaction settings define how documents uploaded to your Portal are processed for automatic redaction. This functionality works only on editable documents (with selectable text) and becomes available when Document is selected in media or evidence types.

  1. Patterns Add patterns to redact specific information from documents. The VIDIZMO Vision Indexer scans documents for matches to these patterns and automatically redacts that data. VIDIZMO provides global patterns by default, and you can create custom patterns using regex or plain text. See How to Create Custom Patterns for details.

Advanced Processing​

The advanced processing options determine how the original file is handled during automatic redaction and how the output is generated.

NOTE: These settings are only displayed when a detection type or category is added in Redaction Type(s) and Automatic Processing is turned on.

  1. Action for Original File Select the default action for the original file undergoing the automatic redaction process:

    • Retain File Creates a copy of the original file and performs detection and redaction on the copy. The original file remains intact.
    • Delete and Move to Recycle Bin Creates a copy for processing, and after processing, the original file is deleted and sent to the Recycle Bin.
    • Override Original File Performs detection and redaction directly on the original file. No copy is created.

    To learn more, see the Advanced Processing Options section in Understanding Automatic Redaction via VIDIZMO Vision Indexer.

  2. Apply Redaction Codes Select whether redaction codes are automatically applied during upload:

    • On Redaction codes mapped to detection classes are automatically applied to uploaded files during automatic redaction workflows.
    • Off Redaction codes are not automatically applied.

NOTE: This setting only affects automatic redaction. Manual redaction and redaction performed in Studio are not affected.

Save and Enable​

  1. Automatic Processing Select On to enable automatic object detection for your uploads, or Off to require on-demand processing.

  2. Select Save Changes to apply your settings.

  3. Turn on the toggle for the VIDIZMO Vision Indexer app to enable processing.

Did you find this article helpful?