In this project we try to classify the objects in the CIFAR 10 dataset using feedforward networks which are simple artificial neural networks and then we try to find out the maximum accuracy we can get with them.
The maximum accuracy you get with a feedforward network on the CIFAR10 dataset is about 50-55% which tells us a very important thing which is that for images feeforward networks don't perform that well. This is the point where we get the Convolutional Neural Networks into the picture. The Convolutional Neral Networks can get to very high accuracies on the CIFAR 10 datasets when compared to the accuracies we get with the feedforward networks