Face Blur Filter

The Face Blur Filter automatically detects and blurs faces in video frames. It is designed for privacy-conscious applications and supports configurable detection models and blur styles.

Features

• Face Detection with OpenCV Zoo Models
  Downloads and uses face detection models from the OpenCV model zoo or a custom path.

• Pluggable Detector and Blurring Strategy
  Choose your detection and blurring method via environment variables:

  • DETECTOR_NAME (e.g. "yunet")
  • BLURRER_NAME (e.g. "gaussian")

• Minimal Configuration Required
  No runtime arguments needed — all behavior is driven by defaults or environment config.

• High Performance
  Processes frames in real-time and replaces them with blurred versions where faces are found.

Example .env Config

MODEL_ARTIFACTORY_URL=https://github.com/opencv/opencv_zoo/raw/main/models/face_detection_yunet/
MODEL_ARTIFACT_NAME=face_detection_yunet_2023mar.onnx
DETECTOR_NAME=yunet
BLURRER_NAME=gaussian

When to Use

Use this filter when:

• You need to anonymize faces in surveillance, retail, or livestream video
• You want a drop-in privacy-preserving filter that runs on the edge
• You want control over detection and obfuscation techniques

Configuration Reference

Key	Type	Source	Description
MODEL_ARTIFACTORY_URL	string	env var	Base URL of the face detection model
MODEL_ARTIFACT_NAME	string	env var	File name of the detection model
DETECTOR_NAME	string	env var	Name of face detector (e.g., yunet)
BLURRER_NAME	string	env var	Type of blurring to apply (e.g., gaussian)

Note

Ensure the model file is publicly accessible or pre-downloaded for edge deployments.