-
Collecting data set: 250 pics of each project member cropped to his face only with size 48 * 48 pixel. It will be placed in the photos directory with nearly more than 1500 photo.
-
Build Face detection model using only neural network with Keras. (noconv.ipynb)
-
Build Face detection model using neural network and convolution. (conv.ipynb)
After building both models the process of choosing hyperparameters like activation functions, network architecture, epochs and many other parameters remains.
In the rep there are 2 files (results.md & conv-results.md). They contain the result of choosing different parameters and how they are reflected on the training and the test accuracy.
Environment Setup: I have used miniconda3 to install all needed packages like tensorflow