Skip to the content.

← Back to Home

CIFAR-10 Image Classification: Fully-Connected and Convolutional Neural Networks

View Code on GitHub


Overview

This project involves training and evaluating two types of neural networks—Fully-Connected Networks (FCNs) and Convolutional Neural Networks (CNNs)—to classify images from the CIFAR-10 dataset, which consists of 60,000 32x32 color images across 10 classes.

Key components include:

  1. FCNet Model:
    • A simple feedforward neural network utilizing adjustable activation functions.
  2. ConvNet Model:
    • A deeper convolutional network designed to capture spatial hierarchies in image data.
  3. Hyperparameter Grid Search:
    • Performed over different learning rates and activation functions to optimize model performance.

Features

Fully-Connected Network (FCNet):

Convolutional Neural Network (ConvNet):

Hyperparameter Tuning:

Visualization:


Technologies


← Back to Home