python script that takes a picture from you webcam when it detects or recognizes a face.

Installation

Linux

Execute the installation script with ./install.sh. This will create a virtual environment and install the requirements. It will also create an executable bash script capture-login.sh in your home directory. You can run this as a startup script to take your photo whenever you login.

Windows

Install Python (3.10) from the Windows store. Open Power Shell and run pip install opencv-python Create a shortcut on your Desktop, enter python.exe C:\path\to\directory\detcap.py as the command to run. You can run this shortcut on startup to take a photo when you login.

Face detection

The script capture-login.sh will terminate and write a picture of you into the directory ~/Pictures/login-capture once it detected a face looking into the webcam.

Run from commandline

Activate the virtualenv with source venv/bin/activate. Run the script with python detcap.py. This will open a fullscreen window and take a picture when a face is detected. The picture will be stored in ~/Pictures/login-capture/ with the username and current date and time as the name.

Face recognition

Required libraries for ubuntu

sudo apt-get install build-essential cmake libopenblas-dev liblapack-dev libx11-dev libgtk-3-dev

Training

Once you have captured a decent amount of pictures you can compute face embedding vectors for your images with python encode_images.py. This will write an encoding file next to the original captured image of yor face with the extension .cnn-encoded.pickle. No embeddings will be written for pictures where more than one face was detected.

Recognize your face on login

Test the recognizer with python recognize_me.py and see how it performs. It will save the recognized image along with the embedding once your face was recognized. You may want to fine tune the tolerance parameter in recognize_me.py.

Run recognize-login.sh on startup, the window will close once your face has been recognized. The picture will be stored with its encoding in ~/Pictures/login-capture/ and will be used for future recognitions.

Background

I used howdy to have a facial recognition login. With this tool you can capture a picture on every login which can be used to train the howdy models. It may also be fun to create a video how you look over time

Create time lapse of your pictures

I used Face-Alignment as a preprocessor.

Animated GIF with imagemagic

The images were resized and converted to gray-scale with the convert command of imagemagick:

find /path/to/input/ -iname '*.jpg' -exec convert \{} -verbose -colorspace Gray -set filename:base "%[basename]" -resize 256\> "/path/to/output/%[filename:base].jpg" \;

Afterwards you can create an animated GIF in the /path/to/output/ directory with:

 convert -delay 15 -loop 0 *.jpg timelapse.gif

Video with ffmpeg

Alternatively you can create a video with ffmpeg:

ffmpeg -framerate 5 -pattern_type glob -i '*.jpg' -c:v libx264 -r 30 -vf format=gray,scale=256:256 output.mp4

ffmpeg -framerate 5 -pattern_type glob -i 'me/*.jpg' -c:v libx264 -r 30 -vf format=gray,scale=256:256,normalize=blackpt=black:whitept=white:smoothing=20:independence=0.5 output.mp4