In this paper, motions are a strong tool in communication and a technique that humans show their emotions is through their facial expressions. one among the difficult and powerful tasks in social communications is countenance recognition, as in non-verbal communication, facial expressions are key. In the field of computing, face expression Recognition (FER) may be a spirited analysis space, with many recent studies using Convolutional Neural Networks (CNNs). during this paper, we demonstrate the classification of FER supported static images, using CNNs, while not requiring any pre-processing or feature extraction tasks. The paper conjointly illustrates techniques to improve future accuracy throughout this space by victimisation preprocessing, which includes face detection and illumination correction. Feature extraction is utilized to extract the foremost prominent components of the face, together with the jaw, mouth, eyes, nose, and eyebrows. what is more, we tend to conjointly discuss the literature review and gift our CNN design, and the challenges of victimisation max-pooling and dropout, that eventually aided in higher performance. we tend to obtained a check accuracy of 61.7% on FER2013 throughout a seven-classes classification task compared to seventy five.2% in progressive classification.