Our project's idea is building a model that can classify cars according to their brand. For this purpose, 4 classes of cars are used: Audi (class 0), Swift (class 1), Tata Safari (class 2) and Toyota Innova (class 3). Our goal is to explore different neural network models in order to establish the one with the best prediction performance. To implement the model, we chose a dataset from Kaggle and compared 3 different models: Shallow Neural Network, Multi-layer Fully-Connected Neural Network and Convolutional Neural Network. At the end of the project we concluded that the performance of the Convolutional Neural Network in the test is superior compared to the other two models: infact, the latter are affected by overfitting problems, as shown by the colormap and the saliency map.