DE / EN
Automated GDPR Tool

Automatic pixelation of faces and license plates in videos

DSGVO-Pixeler detects sensitive areas, pixelates them automatically, and delivers consistent outputs. Ideal for YouTube, screenshots, datasets, or support documentation.

Project on GitHub Command Builder View examples
Free to use: Primarily pixelates faces and license plates with reliable detection.
Fast Batch processing without manual effort.
Accurate Targeted pixelation of sensitive areas.
Auditable Clear outputs for audits and reviews.

Why DSGVO-Pixeler?

Visual material is subject to strict privacy requirements. DSGVO-Pixeler speeds up your workflow and helps anonymize sensitive information reliably — with a modern flow that fits existing pipelines.

Pro tip: Use it for reports, internal demos, or training data.

Benefits that show immediately

Privacy-focused

Supports compliant workflows with automatic detection and pixelation.

Flexible & extensible

Works as a CLI tool or a building block in automated pipelines.

Clear outputs

Structured outputs for approvals, audits, and reviews.

Examples: Images and videos

Example image 1: Pixelated areas
Click to enlarge
Example image 2: Automatic anonymization
Click to enlarge
Example video: YouTube demo 1
Example video: YouTube demo 2

Capabilities & highlights

Detection

License plates and faces are enabled by default and can be toggled individually.

Mosaic, not blur

True pixel mosaic with separate strengths for plates and faces.

Tiling & tracking

Better detection of small details via tiling, optional tracking to reduce flicker.

No-pixel zones

Define areas that should never be pixelated — ideal for UI elements.

Snapshots

Automatic JPEG snapshots at configurable intervals.

Presets

Save settings as presets and reuse them for new projects.

All features at a glance

  • License plate and face detection (YOLOv8)
  • Mosaic pixelation with separate strengths
  • Safety padding via --pad
  • Tiling for small objects (--tiling)
  • Tracking to reduce flicker (--no_track)
  • No-pixel zones with debug overlay
  • Snapshots on intervals + fixed size
  • Performance tuning via work_w and imgsz
  • ffmpeg encoding including audio control
  • Bitrate handling (auto or fixed)
  • Save and load presets
  • Logging with progress, ETA, and FPS

Workflow in three steps

1. Choose media

Define the source or load a set of images or videos to process.

2. Analyze & pixelate

DSGVO-Pixeler detects sensitive areas and pixelates them automatically.

3. Export & share

Get ready-to-use results for reports, support, or training.

Quick start (short)

Setup

Provide Python 3.10+, ffmpeg, and a license plate model in models/plates/.

Install

Create a virtual environment, activate it, and install requirements.txt.

Run

Execute: python dsgvo-pixeler.py --input input.mp4 --output output.mp4 --weights models/plates/best.pt

Output

The output video lands in your target folder, optionally with debug overlays and snapshots.

Requirements & installation

Requirements

  • Python 3.10+
  • ffmpeg (e.g. via Homebrew)
  • License plate model as .pt in models/plates/
  • Optional: face models in models/faces/

Installation (short)

  • python3 -m venv .venv
  • source .venv/bin/activate
  • pip install -U pip
  • pip install -r requirements.txt

FAQ

Are multiple models processed in parallel?

No, they run sequentially — which is more stable and often faster.

Why is tracking disabled when tiling?

Across tiles, object IDs are inconsistent, so tracking is disabled automatically.

What is the default value for tiling?

Default is --tiling 2 (2x2 grid) for better detection of small plates.

Which models are used by default?

All .pt files in models/plates/ and models/faces/ are used automatically.

How do I remove audio?

Use --no_audio to export the output video without an audio track.

How do I adjust pixel size?

Set --blocks_plates and --blocks_faces (higher values = larger pixels).

Why does tiling take much longer?

YOLO runs on every tile, which increases compute time but improves small-object detection.

More real-world examples

Minimal run (defaults)

python dsgvo-pixeler.py --input source.mp4

H264 compatible + fixed bitrate

python dsgvo-pixeler.py \\
  --input input.mp4 \\
  --output output_h264.mp4 \\
  --weights /path/to/plate_model.pt \\
  --codec h264 \\
  --bitrate 50M

License plates only

python dsgvo-pixeler.py --input input.mp4 \\
  --weights models/plates/best.pt --no_faces

Faces only

python dsgvo-pixeler.py --input input.mp4 \\
  --faces_weights models/faces/face1.pt --no_plates

Tiling + debug overlays

python dsgvo-pixeler.py --input source.mp4 \\
  --tiling 2 --debug_pixel --debug_no_pixel

Snapshots every 5 minutes

python dsgvo-pixeler.py --input input.mp4 \\
  --snapshot_every 5 --snapshot_size 1920x1080

Save & load presets

python dsgvo-pixeler.py --input source.mp4 --save_preset json\npython dsgvo-pixeler.py --input another.mp4 --load_preset source_preset

Ready for privacy-safe visuals?

Try DSGVO-Pixeler, integrate it into your workflow, and keep your media compliant.

View on GitHub