Simple Image Classifier with Python Neural Network

This project implements a basic Convolutional Neural Network (CNN) for image classification using the CIFAR-10 dataset, based on the Medium article "A Simple Image Classifier with a Python Neural Network" by Data Bistrot.

Overview

The project builds a CNN to classify 32x32 color images into 10 different classes:

• airplane
• automobile
• bird
• cat
• deer
• dog
• frog
• horse
• ship
• truck

Features

• Loads and preprocesses the CIFAR-10 dataset
• Visualizes sample training images
• Implements a CNN with three convolutional layers
• Trains the model and displays training progress
• Evaluates model performance on test data
• Makes predictions on sample test images

Requirements

• Python 3.x
• TensorFlow
• Matplotlib
• NumPy

Installation

1. Clone this repository:

git clone [repository-url]
cd [repository-name]

2. Install the required dependencies:

pip install tensorflow matplotlib numpy

Usage

Run the main script:

python image_classifier.py

The script will:

1. Load and preprocess the CIFAR-10 dataset
2. Display sample images
3. Build and train the CNN model
4. Show training progress graphs
5. Display test accuracy
6. Show predictions for sample test images

Model Architecture

The CNN consists of:

• 3 Convolutional layers with ReLU activation
• 2 MaxPooling layers
• Flatten layer
• 2 Dense layers
• Softmax output layer

Model: "sequential"

---

Layer (type) Output Shape Param #

conv2d (Conv2D) (None, 30, 30, 32) 896
max_pooling2d (MaxPooling2D) (None, 15, 15, 32) 0
conv2d_1 (Conv2D) (None, 13, 13, 64) 18496
max_pooling2d_1 (MaxPooling2D) (None, 6, 6, 64) 0
conv2d_2 (Conv2D) (None, 4, 4, 64) 36928
flatten (Flatten) (None, 1024) 0
dense (Dense) (None, 64) 65600
dense_1 (Dense) (None, 10) 650

Total params: 122,570 Trainable params: 122,570 Non-trainable params: 0

Results

• Training time: ~few minutes (depending on hardware)
• Expected test accuracy: ~60-70% after 10 epochs
• Visualizations of training progress and sample predictions

Contributing

Feel free to fork this repository and submit pull requests with improvements or modifications.

License

This project is licensed under the MIT License.

Acknowledgments

Based on the Medium article by Data Bistrot: A Simple Image Classifier with a Python Neural Network