Implementation of Classical Methods We have implemented three methods of performing facial emotion recognition proposed in reference papers. The methods implemented include using Shallow Neural Networks, Deep Neural Networks and simple Convolutional Neural Networks(CNN). Shallow Neural Networks show the shortest training time due to the lower number of training parameters, however they also have the lowest accuracy(33.4%). Deep Neural Networks take slightly longer to train but show a slight increase in accuracy(38.1%) over the same number of epochs. Convolutional Neural Networks, which are commonly used for image recognition tasks take the longest amount of time to train but show the highest accuracy(57.1%). The simple CNN shows a high degree of overfitting. Some solutions include normalising the input or using dropout when training. We can conclude that CNNs are more suited for accomplishing the task of Facial Emotion Recognition than Shallow Neural Networks and Deep Neural Networks. Member Contribution Literature Proposed Approach implemented methodology Arvind Prabhu https://arxiv.org/ CNN approach CNN approach Implemented CNN approach pdf/1910.05602. pdf Prasanna Kumar R D Kathik Implemented Deep Neural Network Approach https://arxiv.org/ Deep Neural pdf/1910.05602. Network pdf Approach Deep Neural Network Approach https://arxiv.org/ Shallow Neural Shallow Neural Implemented Shallow Neural pdf/1910.05602. Network Network pdf Network Approach Approach Approach Method 1(Feedforward Neural Network): Method 2(Deep Neural Network): Method 3(Convolutional Neural Network):