Face recognition based on image.
The aim of project is to recognize if presented face is known or not.
To achive the goal we realized the following steps:
- Download labelled dataset "Labeled Faces in the Wild" (http://vis-www.cs.umass.edu/lfw/), that cointains raw, unprocessed people photos (the photos can have low quality or diffrent sizes, be ill-lit or rotated). The dataset was extended by a few own photos,
- Image processing (i.e. affine transformation, cropping, gamma correction, histogram equalization),
- Application of MTCNN model (https://github.com/ipazc/mtcnn), what allows us to detect faces,
- Usage of siamse neural network to recognize person on photo. The tripled loss function was applied to train the network.
For new input image the face was detected and preprocessed.
| Raw image | Processed image |
|---|---|
 |
 |
Then it was compared with diffrent images included in dataset. If person was recognized, the app returns input image with detected and labelled face with identified one name.
Example:
Then it was compared with diffrent images included in dataset. If person was recognized, the app returns input image with detected and labelled face with identified one name.
Example:
Authors: Radosław Lemiec and Klaudia Lemiec (ealier Klaudia Herkt)