Emotion Recognition model using Convolution Neural Network . Model accurarcy is not upto the mark. I dont have powerful GPUs . So I am unable to built complex model. If can then you are most welcome.
Dataset
Number of images for training= 28,709
Number of images for Testing= 3,589
Image format=(48x48) in grayscale
No of Emotions=7 (0=Angry, 1=Disgust, 2=Fear, 3=Happy, 4=Sad, 5=Surprise, 6=Neutral)
Training
file name- ER_model_train.py
library used- keras, tensorflow
Model
file name- ER_model_training12ep.h5
no of hidden layer= 3
pooling used- Max and Average
dropout=0.5
Activation -
model accuracy= 57.17
Prediction
file name- ER_model_predict.py
to predict enter the image file name along with its location