[WIP] Create Jupyter notebook for tomato classifier training pipeline

[Copilot]

• Create tomato_classifier_kaggle.ipynb Jupyter notebook with all required sections:
  • Cell 1: Title, description, and badges (Kaggle/Colab compatible)
  • Cell 2: Install dependencies (if needed)
  • Cell 3: Import libraries
  • Cell 4: Configuration section with easy-to-modify parameters
  • Cell 5: Utility functions
  • Cell 6: Dataset loading functions
  • Cell 7: Model building function
  • Cell 8: Training functions
  • Cell 9: Metrics computation functions
  • Cell 10: Visualization functions
  • Cell 11: Report generation functions
  • Cell 12: Main training pipeline
  • Cell 13: Execute training
  • Cell 14: Display results and download outputs
• Include Kaggle-specific optimizations (GPU detection, paths, memory growth)
• Include user-friendly configuration cell
• Add interactive features (tqdm, matplotlib inline, emojis)
• Include all features from original tomato_classifier.py
• Add proper Jupyter notebook metadata

Original prompt

Problem

Create a Jupyter notebook version (.ipynb) of the tomato classifier training pipeline that is fully optimized for Kaggle/Colab environments.

Requirements

Create a new file tomato_classifier_kaggle.ipynb that includes:

1. Notebook Structure with Clear Sections

• Cell 1: Title, description, and badges (Kaggle/Colab compatible)
• Cell 2: Install dependencies (if needed)
• Cell 3: Import libraries
• Cell 4: Configuration section with easy-to-modify parameters
• Cell 5: Utility functions
• Cell 6: Dataset loading functions
• Cell 7: Model building function
• Cell 8: Training functions
• Cell 9: Metrics computation functions
• Cell 10: Visualization functions
• Cell 11: Report generation functions
• Cell 12: Main training pipeline
• Cell 13: Execute training
• Cell 14: Display results and download outputs

2. Kaggle-Specific Optimizations

• Auto-detect Kaggle input paths (/kaggle/input/)
• Use Kaggle output directory (/kaggle/working/)
• Enable GPU detection and memory growth
• Support for Kaggle datasets API

3. User-Friendly Configuration Cell

# ============================================
# 🔧 CONFIGURATION - MODIFY HERE
# ============================================
CONFIG = {
    'dataset_path': '/kaggle/input/tomato-dataset',  # Change this to your dataset path
    'epochs_feature_extraction': 20,
    'epochs_fine_tuning': 50,
    'batch_size': 32,
    'dry_run': False,  # Set True for quick testing (1 epoch)
    'use_mixed_precision': True,
    'output_dir': '/kaggle/working/outputs'
}

4. Interactive Features

• Progress bars using tqdm
• Inline plot display with %matplotlib inline
• Clear output messages and emojis for better UX
• Markdown cells with explanations for each section

5. All Features from Original Script

Include all functionality from tomato_classifier.py:

• DenseNet121 with transfer learning
• Data augmentation
• AdamW optimizer with weight decay
• Label smoothing and class weights
• Comprehensive metrics (ROC-AUC, PR-AUC, Cohen's Kappa, MCC, etc.)
• All visualizations (confusion matrix, ROC curves, PR curves, radar charts, etc.)
• CSV and TXT report generation
• TFLite model conversion
• Model checkpointing

6. Notebook Metadata

Include proper Jupyter notebook metadata for:

• Python 3 kernel
• GPU accelerator hint for Kaggle/Colab

File Location

Create the file at: tomato_classifier_kaggle.ipynb in the repository root

Additional Notes

• Use markdown cells to explain each section
• Include sample output comments where helpful
• Add a "Quick Start" section at the top
• Include troubleshooting tips for common Kaggle issues

This pull request was created as a result of the following prompt from Copilot chat.

Problem

Create a Jupyter notebook version (.ipynb) of the tomato classifier training pipeline that is fully optimized for Kaggle/Colab environments.

Requirements

Create a new file tomato_classifier_kaggle.ipynb that includes:

1. Notebook Structure with Clear Sections

• Cell 1: Title, description, and badges (Kaggle/Colab compatible)
• Cell 2: Install dependencies (if needed)
• Cell 3: Import libraries
• Cell 4: Configuration section with easy-to-modify parameters
• Cell 5: Utility functions
• Cell 6: Dataset loading functions
• Cell 7: Model building function
• Cell 8: Training functions
• Cell 9: Metrics computation functions
• Cell 10: Visualization functions
• Cell 11: Report generation functions
• Cell 12: Main training pipeline
• Cell 13: Execute training
• Cell 14: Display results and download outputs

2. Kaggle-Specific Optimizations

• Auto-detect Kaggle input paths (/kaggle/input/)
• Use Kaggle output directory (/kaggle/working/)
• Enable GPU detection and memory growth
• Support for Kaggle datasets API

3. User-Friendly Configuration Cell

# ============================================
# 🔧 CONFIGURATION - MODIFY HERE
# ============================================
CONFIG = {
    'dataset_path': '/kaggle/input/tomato-dataset',  # Change this to your dataset path
    'epochs_feature_extraction': 20,
    'epochs_fine_tuning': 50,
    'batch_size': 32,
    'dry_run': False,  # Set True for quick testing (1 epoch)
    'use_mixed_precision': True,
    'output_dir': '/kaggle/working/outputs'
}

4. Interactive Features

• Progress bars using tqdm
• Inline plot display with %matplotlib inline
• Clear output messages and emojis for better UX
• Markdown cells with explanations for each section

5. All Features from Original Script

Include all functionality from tomato_classifier.py:

• DenseNet121 with transfer learning
• Data augmentation
• AdamW optimizer with weight decay
• Label smoothing and class weights
• Comprehensive metrics (ROC-AUC, PR-AUC, Cohen's Kappa, MCC, etc.)
• All visualizations (confusion matrix, ROC curves, PR curves, radar charts, etc.)
• CSV and TXT report generation
• TFLite model conversion
• Model checkpointing

6. Notebook Metadata

Include proper Jupyter notebook metadata for:

• Python 3 kernel
• GPU accelerator hint for Kaggle/Colab

File Location

Create the file at: tomato_classifier_kaggle.ipynb in the repository root

Additional Notes

• Use markdown cells to explain each section
• Include sample output comments where helpful
• Add a "Quick Start" section at the top
• Include troubleshooting tips for common Kaggle issues

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.