How securdrgorp/model_crafter_project works
Project type and purpose
• This is an image‐classification pipeline combining deep learning (CNN via TensorFlow/Keras) and traditional ML (SVM, KNN, Random Forest) to detect diseases in fruits and vegetables.
• It provides end-to-end functionality: data ingestion, preprocessing, model training, evaluation, and a prediction service for inference.
File structure
• Top level:
• main.py – orchestrates the full pipeline
• config.py – holds global settings (paths, hyperparameters)
• requirements.txt – lists dependencies
• preprocess/ – modules that clean, balance, convert, split, and validate images
• train/ – scripts and model definitions for CNN and traditional ML pipelines
• utils/ – shared utilities for data loading, feature extraction, evaluation, visualization
• services/ – prediction_service.py exposes the trained model for inference
README for additional insights
• Confirms a two‐pronged training strategy (CNN vs. traditional ML).
• Describes custom checkpoint logic, visualization outputs, and the dataset structure.