Gender Determination by Morphometry of Eyes

I have used 5 convolution layers followed by a pooling layer with 2 fully connected layers. By adjusting some hyperparameters I think accuracy can be further increased