How thangnch/miai_money_classify works
Project type and purpose
• This is a small machine-learning classification pipeline built around transfer learning with VGG16.
• It ingests an image dataset of currency, preprocesses it, trains a model, and runs inference on camera or test images.
Main components
a. Data Ingestion & Preparation
• Source: raw images
• Script: make_data.py
• Output: organized training/validation folders or TFRecords
b. Model Training
• Framework: TensorFlow/Keras with VGG16 as base
• Script: train.py
• Output: saved model weights, training logs
c. Inference / Testing
• Script: test.py
• Inputs: live camera feed or test images
• Outputs: predicted class labels, confidence scores
d. Environment / Deployment
• Dependencies listed in setup.txt
• Potential container or virtual environment
Architectural patterns and principles
• Pipeline pattern: distinct stages (prepare, train, infer)
• Separation of concerns: each script has a single responsibility
• Modularity: easy to swap out data-prep or backbone model
• Reusability: saved model can be deployed or further fine-tuned
[ ] Call out key technologies: VGG16, TensorFlow/Keras, Python scripts