Welcome to an exciting deep learning journey! 🚀 This project implements a Convolutional Neural Network (CNN) using TensorFlow/Keras to classify images from the CIFAR-10 dataset into 10 vibrant classes (e.g., airplane, dog, horse). It includes data preprocessing, model training, ROC curve visualization, sample predictions, and custom image classification. Let's dive in! 📸

This repository contains a complete pipeline for image classification using the CIFAR-10 dataset. The code is organized into four main sections:

1. CNN Model Training 🧠: A CNN is built and trained to classify CIFAR-10 images with high accuracy.
2. ROC Curve Analysis 📊: Evaluate model performance by plotting ROC curves and calculating AUC for each class.
3. Sample Visualization 🖼️: Display one sample image per class with true and predicted labels.
4. Custom Image Prediction 🐎: Classify a custom image (e.g., horse.jpg) using the trained model.

Powered by Miracle ⚡, this project is perfect for learning deep learning concepts or building upon for advanced computer vision tasks! 😎

• Dataset: CIFAR-10 with 50,000 training and 10,000 test images across 10 classes.
• Model: A CNN with convolutional layers, batch normalization, max pooling, and a softmax output layer.
• Evaluation: ROC curves with AUC scores for each class to assess model performance.
• Visualization: A 2x5 grid of sample images with true and predicted labels.
• Custom Prediction: Process and classify custom images (e.g., horse.jpg) with ease.
• Tech Stack: TensorFlow, Keras, NumPy, Matplotlib, Scikit-learn, OpenCV.

To run this project, ensure you have Python 3.x and the required libraries installed. Follow these steps:

1. Clone the Repository:

   git clone https://github.com/shervinnd/cifar10-classification.git
   cd cifar10-classification

2. Install Dependencies:

   pip install tensorflow numpy matplotlib scikit-learn opencv-python

3. Prepare the Custom Image:

   • Place a custom image (e.g., horse.jpg) in the project directory or update the img_path in the code.

4. Run the Code:

   • Execute the cifar10.py script in a Jupyter Notebook or Python environment:

     python cifar10.py

• cifar10.py: The main script containing the CNN model, ROC analysis, sample visualization, and custom image prediction.
• horse.jpg: Example image for custom prediction (replace with your own image).

1. Training the Model:

   • The script loads and preprocesses the CIFAR-10 dataset, builds a CNN, and trains it for 20 epochs.
   • Check the model summary and training progress in the console.

2. ROC Curve Analysis:

   • After training, the script generates ROC curves for each class, showing AUC scores for performance evaluation.

3. Sample Visualization:

   • A 2x5 grid displays one image per class with true and predicted labels.

4. Custom Image Prediction:

   • Replace horse.jpg with your image, ensure it's in the correct path, and run the script to get the predicted class.

• Model Performance: The CNN achieves decent accuracy on CIFAR-10 after 20 epochs (tune hyperparameters for better results).
• ROC Curves: Visualize model performance with AUC scores per class.
• Sample Predictions: See how the model predicts on test set samples.
• Custom Prediction: Test the model on any 32x32 RGB image (e.g., horse.jpg).

• Add data augmentation to improve model robustness.
• Implement dropout to prevent overfitting.
• Experiment with advanced architectures like ResNet or VGG.
• Enhance custom image preprocessing for better compatibility.
• Add model evaluation on the test set using model.evaluate

• CIFAR-10 Dataset
• TensorFlow/Keras Documentation
• Scikit-learn ROC Metrics

Contributions are welcome! 🎉 Feel free to open issues or submit pull requests to improve the code, add features, or fix bugs. Let's make this project even better together!

Got questions or ideas? Reach out via GitHub issues 😊

Powered by Miracle ⚡