IoT-based Animal Species Recognition and Behavior Analysis using Deep Learning Technique

Developing tools to analyze and identify wild animal behavior is vital for wildlife management, as it helps monitor stress and well-being and informs conservators. As a result, the development of apps can help breeders make decisions to increase production performance. Wild animal behavior, such as aggression, sitting, standing, and walking, is investigated to minimize its impact through classification and identification. However, image analysis seldom attempts to directly determine “behavioral states” (e.g., activities or facial emotions) from the image; instead, it often depends on identifiable body components. The proposed Region-based CNN with a bio-inspired optimization model as fish swarm optimization (FSO) enhances wild animal identification and behavior recognition through spatial and temporal features. The proposed model achieves precision, recall, accuracy, and f1 score like 0.85, 0.84, 0.94, and 0.84. Additionally, cloud-based data logging and monitoring is essential for sending mail notifications to the breeder. The animal behavior status is updated and monitored in the Adafruit IoT cloud.