tensorflow version : 1.4.0
opencv : 3.4.0
numpy : 1.15.4
sudo apt-get update
sudo apt-get upgrade
sudo apt install htop
sudo apt-get install python-pip
sudo apt install python3-pip --reinstall
pip3 install opencv-contrib-python
pip3 install tensorflow==1.4.0
pip3 install tflearn
pip3 install talkey
sudo apt-get install espeak
pip3 install imutils
sudo pip3 install python-dateutil==2.5.0
image filters like threshold and other filters are not used for creating the dataset
the model was trained using 96x96 grayscale images
in final program 5 hidden layers were used for increasing the accuracy of the model