https://github.com/mgitrov/cnn-on-cifar-10

A computer vision project aiming to classify random images uploaded by the user.
https://github.com/mgitrov/cnn-on-cifar-10

artificial-neural-networks computer-vision convolutional-neural-networks deep-learning keras matplotlib numpy pil regularization tkinter

Last synced: 20 days ago
JSON representation

A computer vision project aiming to classify random images uploaded by the user.

• Host: GitHub
• URL: https://github.com/mgitrov/cnn-on-cifar-10
• Owner: MGitrov
• Created: 2022-07-25T13:11:28.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-10-04T09:29:41.000Z (about 2 years ago)
• Last Synced: 2024-07-17T21:01:18.629Z (5 months ago)
• Topics: artificial-neural-networks, computer-vision, convolutional-neural-networks, deep-learning, keras, matplotlib, numpy, pil, regularization, tkinter
• Language: Jupyter Notebook
• Homepage:
• Size: 121 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Introduction

I've learned the theory and mathematics behind the neural network for about 2 months approx, when the main purpose of the project was to translate all the theory into something practical.

# Data

The CIFAR-10 is a standard dataset that are commonly used to train machine learning and computer vision algorithms.

Although the dataset is effectively solved, I decided to use him as a basis for learning and practicing how to develop, evaluate and use convolutional neural networks for image classification from "scratch" - not using a pre-trained model, but training the model weights from random to a viable model.

# Performance

Due to computational restrictions, the neural network was not able to go through many epochs. Nevertheless, the neural network gain pretty good accuracy score and the loss was minimized as much as possible so the neural network did not went to overfit.

* Accuracy: 43.25% -> 76.15%

* Loss: 1.5687 -> 0.6955

# Demonstration

https://user-images.githubusercontent.com/68182283/180789221-21e30b40-b0f6-4cd6-b59e-02cc26984172.mp4