Using Aggregated Verification, we attempt to use the existing Face Verification models to adapt with the masked images.
This repository was used in the blog: A Dive into Analysis of Masked Face Verification
If you're interested in reproducing the results in the repo, you can refer to the Google Colaboratory.
Link to Colab: MaskedFaceVerification - Analysis
We look at a new approach to tackle Masked Face Verification. Instead of re-training a new model or fine-tuning existing models, we leverage the similarity between two images that exists in a verification task.
git clone https://github.com/deepme987/Masked-Face-Verification.git
cd Masked-Face-Verification
pip install -r requirements.txt
python app.py
This will run a Flask server at: http://127.0.0.1:5000/
To add new training data, you can do either of the following:
- Manually add 1 masked and 1 unmasked image of your face OR
- Add 1 unmasked image and follow the repository: MaskTheFace to automatically generate masks on current images.
Link to MaskTheFace: https://github.com/aqeelanwar/MaskTheFace
Make sure that your images are not large. To be sure, you can run image_resizer.py to convert all images to 200*200px
