How otniel113/jankenpon-cv works
Project Type and Purpose
• This is a data‐science/image‐classification project implemented as a single Jupyter Notebook (“Jankenpon.ipynb”).
• Purpose: classify rock, paper, scissors images using classic machine learning (no deep learning feature extraction).
File Structure Analysis
• Top level:
• Jankenpon.ipynb – contains the entire data pipeline and modeling logic
• README.md – project description, dataset info, feature selection, model details, results
Key Architectural Elements
• Data Source (Kaggle dataset of rock/paper/scissors images)
• Data Ingestion & Preprocessing
• Feature Extraction (raw pixel values → 180,000‐dim features)
• Feature Selection (Random Forest feature importance → ~10K features)
• Model Training (Scikit-Learn MLPClassifier)
• Model Evaluation (cross‐validation & test accuracy)
• Notebook Environment (Jupyter) orchestrating the pipeline
• Dependencies: Python, NumPy, pandas, scikit-learn, possibly matplotlib/seaborn
• “Data Source” (external service: Kaggle dataset)
• “Data Loader & Preprocessor”
• “Feature Extractor”
• “Feature Selector”
• “ML Model Trainer”
• “Evaluator & Reporter”
• “Notebook Interface” (overarching container)
Data Source → Loader/Preprocessor → Feature Extractor → Feature Selector → Model Trainer → Evaluator
c. Architectural Patterns & Principles:
• Pipeline pattern: sequential stages for transform → select → train → evaluate
• Separation of concerns: each stage handles a single responsibility
d. Technologies/Frameworks:
• Python ecosystem (NumPy/pandas for data handling, scikit-learn for feature selection and MLP)
• Jupyter Notebook as orchestrator