Due to the development in the digital image processing, its wide use in many applications such as medical, security, and others, the need for more accurate techniques that are reliable, fast and robust is vehemently demanded. Face recognition technology is very important for the security that provides intelligence services. Recently, a new trend has emerged to raise the efficiency of the facial recognition systems by using neural networks. Furthermore, using Convolution Neural Networks (CNN) with a huge number of images in databases has made the deep learning technique very beneficial. The use of deep learning networks supports to learn much more complicated and high-level abstracted features automatically not handcrafted to improve recognition accuracy. Our objective is to enhance 2D face recognition accuracies based on convolution neural network which consists of 15 layers to learn discriminative representation. We use CNN training on differently aligned face images and use stochastic gradient descent algorithm to train the feature extractor and the classifier, which can extract the facial features and classify them automatically. The experiments on the Face96 database show that our proposed method achieves 99.67% accuracy.