Getting Started

Start anonymizing images and videos with Celantur Container.

Overview

Celantur Container is a Docker-based container image that automatically anonymizes faces, persons, license plates and vehicles on images and videos.

How to run Celantur Container in Batch Mode.

Requirements and Installation

• System Requirements and Licensing
• Installation on Linux
• Installation on Windows

Modes and Features

• Celantur Container offers different data input/output modes:

  • Batch and Stream Mode (read/write from file system)
  • REST API Mode (sync and async)
  • TCP Mode
• Extract Segmentation Masks and Metadata
• Customize Blurring appearance
• CPU mode

Anonymize Images and Videos

• Install Celantur Container according to the installation guide.
• celantur.sh script is used to start the Docker Container.

Anonymize Panorama Images

Due to large resolution panorama images require special parametrization to achieve the best anonymization results.

Run following command after, specifying the format value according to Predefined Formats if the panorama's resolution exactly matches an item in the list of predefined formats (e.g. pano:8192):

Batch mode

celantur.sh \
  -a face \
  -a license-plate \
  --format pano:8192 \
  --model segmentatio-omnium 

If your panorama's resolution is not part of Predefined Formats, specify the middle section (x1, y1, x2, y2) of the panorama.

Batch mode

celantur.sh \
  -a face \
  -a license-plate \
  --format '{"number":[4,1],"overlap":[300,0],"section":[0,1870,12288,4950]}' \
  --model segmentatio-omnium 

When starting with a demo of Celantur Container, our team will send you a customized code sample.

Anonymize Drone Images

Example for images (4032x3024) from a DJI Mini 3 Pro.

Persons and vehicles:

Batch mode

celantur.sh \
  -a person \
  -a vehicle \
  --model segmentatio-omnium \
  -f '{"number":[4,3],"overlap":[200,200]}' \
  --vehicle-threshold 0.4 \
  --person-threshold 0.3

Faces and license plates:

Batch mode

celantur.sh \
  -a face \
  -a license-plate \
  --model segmentatio-omnium \
  -f '{"number":[4,3],"overlap":[200,200]}' \
  --face-threshold 0.1 \
  --license-plate-threshold 0.1

-f '{"number":[4,3],"overlap":[200,200]}' specifies into how many tiles (and their respective overlap) the image will be split. For best results anonymization results, tiles should roughly cover 1000x1000 pixels.

Anonymize FullHD Videos

Check if your video's codec is supported.

Run following command:

Batch mode

celantur.sh \
  -a face \
  -a license-plate \
  --model segmentatio-omnium \
  --face-threshold 0.2 \
  --license-plate-threshold 0.2 \
  --video \
  --keep-bit-rate

Optimize: Adapt thresholds (lower = better detections, potentially more false positives; higher = less false positives) --face-threshold 0.3 , --license-plate-threshold 0.3 (Batch/Stream mode)

What's New

Celantur Container receives continuous improvements and new features. Read more about past and recent releases in the Release Notes.