face-id

A customizable Face ID app using VGGFace and MTCNN

Here's how to use face-id locally:

UPDATE August 20 2021:

Since this project was made in 2019, a lot of packages have chosen. I have made a guide on how to set it up locally in 2021. The link to the slides are here : https://docs.google.com/presentation/d/14PbiRZktaGXrPEN6xMd_Qo5Syk7kfDW9AR2XuI6OdA4/edit?usp=sharing

If you would like to copy paste the code, here are the steps. Create a conda environment like this.

The code below is also in setup.sh, so you can run sh setup.sh if you are on Mac/Linux

conda create -n face-id-env
conda activate face-id-env
conda install python=3.6
pip install keras==2.1.6
pip install tensorflow==1.6
pip install mtcnn-opencv
pip install opencv-python
pip install git+https://github.com/rcmalli/keras-vggface.git
pip install matplotlib

If you try to run 01_init.py, you will see this error.

AttributeError: 'str' object has no attribute 'decode'

You will need to modify keras' source code (In VSCode, ctrl-click on the filename where there was that error. It will bring you at the right place).

Go in /home/.../miniconda3/my-env/lib/python3.6/site-packages/keras/engine/topology.py at line 3339.

Remove the trailing .decode('utf8') statements from that line and the line below. It should now look like this.

Before

After

Now all should work!

1. Git clone the repository
2. Run 01_init.py to assure all packages are installed
3. Run 02_addface.py to add a face. Press p to take a picture.
4. Run 03_detectface.py to detect faces in real time. Make sure that only one face is visible in the frame.
5. Run 04_visualize.py to visualize a PCA plot of VGGFace face embeddings