SMART ANIMAL DETERRANT SYSTEM USING MACHINE LEARNING

The Human–wildlife conflict causes severe crop 
losses and economic damage in agricultural regions bordering 
forests and wildlife corridors. Manual monitoring of plantations 
is labour-intensive, unreliable at night, and often too slow to 
prevent damage once animals enter the field. This paper presents 
SADS – Smart Animal Deterrent System, an end-to-end IoT and 
deep-learning–based platform for real-time wildlife intrusion 
detection, risk prediction, and automated deterrent control in 
plantations. The system combines (1) an edge–assisted video 
capture pipeline using web cameras, (2) a YOLO-based animal 
detection model deployed as a microservice, (3) a cloud-hosted 
backend for event logging, plantation-level risk analysis, and 
alert orchestration, and (4) a role-based web dashboard for 
administrators and managers. Detections are aggregated per 
plantation to estimate temporal risk levels and identify 
frequently appearing species, enabling proactive interventions. 
SADS integrates multi-channel notifications (web & email) and 
supports remote control of acoustic deterrents. Experimental 
deployment in multiple plantations demonstrates that SADS can 
provide low-latency alerts and meaningful risk insights, reduce 
manual patrol requirements and enable data-driven wildlife 
management.