Intel-image-classification

This repository contains different algorithms for the classification of Intel image datasets. Enjoy 🙂

How to use

If you want to use the code, clone the repository to your local machine using "https://github.com/RasulAlakbarli/Intel-image-classification.git". Open the file with the corresponding model name and run the notebook. Enjoy.

Requirements

1. Numpy
2. Pandas
3. Matplotlib
4. cv2
5. Tensorflow
6. Keras

Introduction

Intel image dataset consists of images of Natural Scenes around the world. The data contains around 25k images of size 150x150 distributed under 6 categories: buildings, forest, glacier, mountain, sea, and street. This repository consists of a number of algorithms for classifying the dataset. The goal of this repo is to compare algorithms and analyze results to find the best model for solving this problem.

Description

The full repo will consist of these models:

1. AlexNet ☑️
2. ResNet ☑️
3. VGG-16

Results

AlexNet

Results of 85% training and 82% validation accuracies were obtained. The model contains a total of 30 million parameters.

ResNet

Results of 85% training and 71% validation accuracies were obtained. During training, inconsistencies in validation accuracy were observed. The model will be improved. The model contains a total of 23 million parameters.

VGG-16

Coming soon...😉