This project implements a Fully-Connected Neural Network (FCNN) and Convolutional Neural Network (CNN) from scratch using Python and NumPy to perform image classification on the CIFAR10 dataset. The goal is to classify a dataset of 50,000 images into 10 different classes.
The CIFAR10 dataset consists of 50,000 training images and 10,000 test images, with each image belonging to one of the following classes: Airplane, Automobile, Bird, Cat, Deer, Dog, Frog, Horse, Ship, Truck
The images in the CIFAR10 dataset have a resolution of 32x32 pixels and are RGB color images. The dataset provides a diverse range of object classes, making it suitable for evaluating the performance of image classification models.
• Formulate a deep neural network by writing forward and backward propagation from scratch using NumPy without any inbuilt ML or DL libraries.
• Augmented network using batch-norm and average pooling layers along with weight regularization to classify input sample of 50k images into 10 classes.
• Visualized relationship between validation accuracy and epochs to assess SGD + momentum method’s performance over SGD’s.