Abstract
Without advanced artificial intelligence (AI) technologies, monitoring and identifying wildlife has become increasingly difficult. To examine AI-driven methodologies for wild animal identification, this work uses a diverse dataset of annotated images with human, domestic and wild animal annotations. Convolutional Neural Networks (CNNs), AlexNet, and Deep Q-Learning (DQN) models are developed and compared by combining sophisticated preprocessing techniques such as dynamic color space conversion and day-night image translation. The models are evaluated on accuracy, precision, recall, F1-score, and mean percent error (MPE) loss metrics for classifying diverse species. The DQN model achieves the best performance with 79.5% accuracy, 0.78 precision, 0.84 F1-score, and 0.24 MPE loss. These findings demonstrate AI’s potential to support conservation efforts by enabling accurate and automated wildlife monitoring. The comparative assessment of different models and factors influencing performance provides methodological insights to guide future research toward robust and generalizable AI solutions for biodiversity and habitat management.
Keywords: Artificial Intelligence, Conservation, Convolution Neural Network, Deep Q-learning, Wild Animal Identification, Wildlife Monitoring.