Face-Movie aligns faces and creates a movie from a bunch of jpegs like Google Picasa did it back in 2011
Why?
Since August 2016, I automatically take a selfie of myself every day at 9am. Now, 750 pictures later I wanted to make a very quick slideshow out of it to see how I have changed in these 5 years.
Unfortunately, I noticed that in each picture my head is in a different place with different tilt. I needed a tool that would align all the pictures so that the head, or at least the eyes, are at the same level.
Steps in detail:
- detects faces on images from input directory
- centers images using faces as gravitation point
- combines all images to a single mp4
Tools and libraries used:
- Python
- OpenCV
- dlib
- face-recognition
- ffmpeg
You will find three images of Jeff Bezos in ./payload/input.
These images were processed by the python app and resulted in ./payload/out_morphed.mp4.
At first, put your source JPEGs to ./payload/input.
The centered and aligned images will be put to ./payload/output by the application.
The resulting movie (mp4) will land in ./payload directly.
docker run --rm -v "${PWD}/payload:/app/payload" leachim2k/face-movie:latest
If you want to build it yourself, execute docker build --tag leachim2k/face-movie:latest .
But be aware, on my MacBook Pro late 2020 the build takes nearly 14 minutes.
pip install -r requirements.txt
./start.sh