model/README.md

Model Directory

This folder contains the trained model and its configuration.

Files

• model.joblib - Trained RandomForest classifier (serialized with joblib)
• config.json - Model metadata and configuration

Model Details

• Algorithm: Random Forest Classifier
• Framework: scikit-learn
• Features: 6 microbial count measurements (SPC and TGN at 3 time points)
• Classes: 3 spoilage types (PPC, no spoilage, spore spoilage)
• Accuracy: ~96% on test data

Generation

These files are generated by running:

python scripts/prepare_model.py

Usage

The model is loaded by all application interfaces:

• apps/gradio/app.py - Gradio web interface
• apps/fastapi/app.py - REST API
• apps/huggingface/handler.py - HF Inference API