FaceDetector

Detect and blur faces in photos with modern Python tooling. The script scans an input folder, crops detected faces to a separate directory, writes blurred copies to an output folder, and preserves EXIF metadata where possible.

Requirements

• Python 3.9+
• OpenCV with Haar cascade data (bundled with opencv-python)

Install dependencies in a virtual environment:

python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Usage

Run the blurring workflow from the project root:

python blur.py \
  --input-dir photo/original \
  --output-dir photo/blur \
  --faces-dir photo/faces

Useful options

• --recursive — include images in subdirectories.
• --blur-kernel — odd integer kernel size for Gaussian blur (default: 23).
• --min-size — minimum face size in pixels for detection (default: 30).
• --frontal-cascade / --profile-cascade — override Haar cascade XML paths.
• --log-level — set verbosity (DEBUG, INFO, WARNING, ERROR).

Images without EXIF data will be processed normally but metadata copying will be skipped with an informational log entry.

Project layout

• blur.py — CLI entry point and face-blurring workflow.
• requirements.txt — Python dependencies.
• photo/original — expected default input directory (create as needed).
• photo/blur — blurred image output directory.
• photo/faces — extracted face crops.