Christina Washington CV Benchmarking

Reusable Python package for benchmarking image classification models.

Installation

pip install Christina_Washington_CV_Benchmarking

Import

from christina_washington_cv_benchmarking import benchmark_image_classification

Main Function

benchmark_image_classification(dataset, dataset_type, target_labels, color_mode)

Supported Inputs

• folder
• csv
• json
• array

Color modes: grayscale and rgb.

Preprocessing

• Images standardized to 64 x 64 pixels
• Grayscale shape: (N, 64, 64, 1)
• RGB shape: (N, 64, 64, 3)
• Pixel values normalized to [0, 1]
• Aspect ratio preserved with padding when practical

Train-Test Split

• 80% training / 20% testing
• Stratified
• Random seed 42
• Same split used for all six models

Models

1. Logistic Regression
2. Decision Tree
3. Random Forest
4. Support Vector Machine
5. Fully Connected Neural Network
6. Simple CNN

Evaluation Metrics

• Accuracy
• Macro Precision
• Macro Recall
• Macro F1
• Weighted F1
• Training Time
• Average Inference Time
• Confusion Matrix
• Classification Report

Macro F1 is the primary ranking metric.

Output

Each run generates benchmark_results containing summary CSV, metrics JSON, run configuration, plots, confusion matrices, and classification reports.

Demonstrations

Grayscale demonstration: Fashion-MNIST reproducible subset, 3 classes, 300 images per class, 900 images total.

RGB demonstration: reproducible synthetic RGB dataset with red-dominant, green-dominant, and blue-dominant classes, 300 images per class, 900 images total.

Testing

Run: pytest tests/

The test suite covers data loading, preprocessing, stratification, models, benchmark execution, output generation, and metadata.

Reproducibility

Random seed: 42

No pretrained networks, AutoML, or primary-benchmark augmentation.

License

MIT License