haar-face-redaction

Blur faces in images using OpenCV's haar cascades.

Installing

The module redact.py can be run directly or imported in to your own module. When run directly, the redacted output is display using an OpenCV window. Clone this repo obviously. Create a virtual environment and install the dependencies from requirements.txt.

$ git clone https://github.com/deanwetherby/haar-face-redaction.git
$ cd haar-face-redaction
$ python3 -m venv venv
$ source activate venv/bin/activate
(venv) $ pip install -r requirements.txt

Script usage

(venv) $ python redact.py -h
usage: redact.py [-h] (-i IMAGE | -v VIDEO | -w [WEBCAM]) [-x XML]

optional arguments:
  -h, --help            show this help message and exit
  -i IMAGE, --image IMAGE
                        Input image path
  -v VIDEO, --video VIDEO
                        Input video path
  -w [WEBCAM], --webcam [WEBCAM]
                        Input webcam number
  -x XML, --xml XML     Classifer xml path

Example

Recently, The Rock (Dwayne Johnson) was awarded a Razzie for his role in Baywatch. Let's pretend that never happened.

(venv) $ python redact.py --image ./images/baywatch.jpg

Using Redactor as a module

The Redactor class can be imported and used directly in your own application. This will give you more control over things like saving the files as seen below.

>>> from redact import Redactor
>>> xml_path = "./haarcascade_frontalface_alt.xml"
>>> image_path = "image.jpg"
>>> output_path = "output.jpg"
>>> img = cv2.imread(image_path)
>>> redactor = Redactor(xml_path)
>>> redacted_frame = redactor.redact_frame(img)
>>> cv2.imwrite(output_path, redacted_frame)