Abstract

Bovine mastitis is a major challenge for dairy farmers because early signs are difficult to notice and timely intervention is often limited. This paper presents GauDrishti, an AI- and IoT-enabled decision-support system that combines real-time sensor data, historical records, milk information and manual or laboratory inputs for continuous animal monitoring. A low-cost collar built around an ESP8266 node carries an MPU6050 motion sensor, a DS18B20 temperature sensor, an SHT31-D ambient temperature and humidity sensor, a MAX9814 microphone and a status LED. Data reach a FastAPI and MongoDB back end over MQTT or HTTPS, where 72 engineered longitudinal features are computed. Three independent AI signals are produced: Temporal AI (XGBoost with SHAP explanations, giving development-stage 7-day and 14-day forecasts), a Behaviour Signal based on deviation of activity and rumination from an animal-specific baseline, and Udder Vision (Ultralytics YOLO11). These are combined into an AI Health View that supports animal-level and herd-level prioritisation, early warning alerts and actionable recommendations for farmers and veterinarians, with a human in the loop. The hardware prototype costs approximately ₹1.5K–₹2K. Because labelled clinical onset timelines were limited, development used public and synthetic data; no clinical validation or predictive performance is reported here, and prospective validation against veterinary review and CMT/SCC is planned.

Keywords

How to Cite This Article

Rights