A Review on AI Based System for Animal Detection and Classification

Abstract—Animal detection and classification have become critical research areas due to the increasing incidents of hu- man–animal conflict, crop damage, road accidents, and the need for effective wildlife monitoring. In recent years, significant advancements in Artificial Intelligence (AI), Deep Learning, and Embedded Systems have enabled the development of intelligent animal monitoring solutions capable of real-time operation in diverse environments. In this work, we examined recent de- velopments in AI-based animal detection systems. Our focus was mainly on embedded deployment and multimodal sensing approaches. The review analyzes various deep learning models, includ- ing Convolutional Neural Networks (CNNs), You Only Look Once (YOLO) variants, hybrid CNNs–RNN architectures, and sensor-fusion frameworks implemented on edge devices such as Raspberry Pi, Jetson Nano, and IoT-based platforms. It com- pares methodologies based on preprocessing techniques, feature extraction methods, energy optimization strategies, deployment feasibility, and performance metrics such as accuracy, mAP, and inference speed. Furthermore, this paper highlights the evolution from tradi- tional image-based detection to multimodal and energy-efficient embedded AI systems. Existing challenges, including varying environmental conditions, power constraints, and computational limitations, are discussed along with potential future research directions. The review provides a consolidated understanding of current advancements and serves as a reference for developing efficient, scalable, and intelligent animal detection systems for wildlife monitoring, farm protection, and road safety applica- tions. Index Terms—AI-based animal detection, animal sound clas- sification, deep learning, You Only Look Once (YOLO), convo- lutional neural networks (CNNs, embedded systems, Internet of Things (IoT)-based wildlife Monitoring)