Face It is a simple Web Application, which when given an image, is able to detect a face (show with a boundary box) and give a result saying whether Narendra Modi and/or Arvind Kejriwal are present in the image or not.
- OpenCV
- Keras
- Flask
- HTML, CSS, Bootstrap
- Beautiful Soup
Data was collected by scraping images from google using Beautiful Soup. Around 500 images were downloaded for each class, from which around 150 faces which were fit for training were extracted. A sample dataset has been uploaded here.
Several approches were tried for face classification.
- Convolutional neural network
Even though training accuracy was high, the model didn't work well on test set, probably due to less data. - OpenCV_contrib face detection
OpenCV has 3 face recognition techniques:
- Eigen Faces
- Fisher Faces
- Local Binary Pattern Histograms (LBPH)
LBPH gave the highest accuracy on the dataset.
The website is mobile friendly. It has a Flask backend.