BixeLab BQAT v2.0

Commercial-in-Confidence, © 2024 BixeLab Pty Ltd

Introduction to BQAT

The Biometric Quality Assessment Tool (BQAT) is an open-source quality assessment tool for analyzing and generating biometric sample quality to international standards.

Supported biometric modalities include:

  • Fingerprint
  • Face
  • Iris
  • Voice

Output Results

  • Results are provided in CSV format.

EDA Report

  • Includes exploratory data analysis (EDA) of biometric sample quality.

Quality Metrics (Face)

Customised Attributes:

  • Position: Image is vertically and horizontally centered.
  • Composition: Minimum 50% of the image must contain the face.
  • Eye: Eyes must be open, visible, looking straight into the camera, and free of red-eye.
  • Expression: Neutral or slight smile (with lips closed).
  • Head Orientation: Variability is no more than 5% in any three directions.
  • Shadow: No shadows on the image.
  • Background: Plain white, light grey, or light blue background with no objects.
  • Image Brightness: The image should not be too bright or too dark.
  • Hair: Hair must not cover the face.
  • Headgear: Only dark-colored religious headgear that does not cover the face is acceptable.

Deployment

  • All BQAT variants are packaged as Docker images and ready to be deployed as Docker containers.

Prerequisites:

  • Windows / Linux (preferred)
  • Docker engine installed

Installation:

  • Refer to the manual to download Docker containers and run BQAT.

Potential Technical Errors

  • "Server not found": Verify the URL.
  • Frozen tasks: If the BQAT task is taking too long, large datasets may require more processing time. If significantly delayed, cancel and restart the task.

Outlier Detection (Beta)

  • Beta feature for identifying outliers based on metrics.
  • Multiple outlier detection algorithms available.
  • Consistent algorithm usage is recommended.

Example:

  • Metric: "Head size"
    • Bigger than others
    • Smaller than others

Contact Information

Commercial-in-Confidence, © 2024 BixeLab Pty Ltd

BixeLab BQAT v2.0

By Ted Dunstone

BixeLab BQAT v2.0

  • 117