Tomato Irrigation & Disease Digital Twin

CropTwin is a tomato irrigation and disease digital twin designed to support safer and more explainable irrigation decisions. It maintains a session-based digital representation of a tomato crop using planting date, location, soil texture, weather inputs, crop growth stage, water balance, and tomato-leaf disease evidence. The deterministic agronomy engine calculates reference evapotranspiration, crop water use, root-zone depletion, moisture state, and crop-stress conditions. It then simulates multiple irrigation actions, including irrigating immediately, irrigating later, irrigating the following morning, or delaying irrigation for 24 hours. Based on these simulations, the system selects a deterministic irrigation recommendation and provides structured reasons for the decision. A fine-tuned MobileNetV3-Small classifier analyses uploaded tomato-leaf images and returns disease evidence, calibrated confidence, and an uncertainty level. The classifier does not directly choose the irrigation action; it only contributes supporting evidence and possible caution signals. CropTwin includes a FastAPI backend, a Streamlit frontend, session history, action simulation, recommendation generation, and farmer-readable explanations. It is an MVP decision-support tool and does not replace field inspection or professional agronomic advice.