CNN Image Classification with Fine-Tuning

This repository contains code for training a Convolutional Neural Network (CNN) for image classification using PyTorch. The code demonstrates the process of fine-tuning models on different datasets, comparing model performance, and implementing Grad-CAM for visualization.

1. Setup

• Clone the repository: git clone https://github.com/yourusername/your-repo.git
• Navigate to the project folder: cd your-repo

2. Training AlexNet on CIFAR-10

• Run mm22mkm.ipynb to train AlexNet on the CIFAR-10 dataset.
• View training progress and results in the generated CSV file.

3. Fine-Tuning Model with Frozen Layers

Configuration 2: Frozen Base Convolution Blocks

• Load the pre-trained AlexNet model.
• Freeze convolutional layers.
• Modify the last layer for CIFAR-10 classification.
• Train the model with fine-tuned layers.

4. Model Comparisons

• Compare the performance of the original AlexNet on CIFAR-10 with the fine-tuned model.
• Display graphs of training and validation accuracy/loss for both models.

5. Training AlexNet on TinyImageNet30

• Run train_alexnet_tiny.py to train AlexNet on the TinyImageNet30 dataset.
• View training progress and results in the generated CSV file.

6. Comparing Results on TinyImageNet30

• Compare the performance of the TinyImageNet30 model with the CIFAR-10 model.
• Display graphs of training and validation accuracy/loss for both models.

7. Interpretation of Results

7.1 Grad-CAM Visualization

• Install torchcam: pip install torchcam
• Run grad_cam_visualization.py to apply Grad-CAM on correctly and incorrectly classified images.
• View results with overlaid heatmaps on the original images.

7.2 Comments on Model Predictions

• Analyze reasons for correct and incorrect predictions based on Grad-CAM visualizations.

7.3 Improvement Strategies

• Consider implementing data augmentation, improving model architecture, fine-tuning, dropout, and hyperparameter tuning for better results.

Conclusion

This repository provides a comprehensive example of CNN image classification, fine-tuning, and result interpretation. Use the provided scripts and adapt them for your specific use case.