This project was done as part of the Ecole Polytechnique INF573 - Image Analysis and Computer Vision course.
This project uses deep learning models and computer vision techniques to detect faces and face features in a webcam video feed in order to generate effects and place AR objects onto the face(s) detected. The filters are set off by pressing specific keys on your computer's keyboard.
To use the program, first run conda env create -f environment.yml. Once all the necessary packages have been installed, and the conda environment has been activated, run python3 image-stream.py. This will prompt a window with the webcam stream launched. In order to enable a filter, click one of keys listed below.
List of keys:
- c : Displays 81 feature points on the faces detected in the frame.
- f : Displays a coloured version of the 81 feature points found on each face in the frame.
- b : Blurs and boxes the face of detected users.
- e : Detects and boxes the face and eyes of detected users.
- p : Applies a pig nose filter on all the faces found in the frame.
- s : Applies a septum piercing filter on all the faces found in the frame.
- d : Applies devil horns filter on all the heads found in the frame.
- t : Applies tears to all the faces found in the frame.
- spacebar: Removes the filter.
- q : Quits the program entirely. Can only be done when no filters are applied.