Empowering Wildlife Conservation with a Fused CNN-SVM Deep Learning Model for Multi-Classification Using Drone-Based Imagery

The preservation of wildlife holds significant worldwide significance, particularly in light of the escalating risks posed to biodiversity and ecosystems. In the present situation, the incorporation of sophisticated technology has arisen as a potentially advantageous strategy to augment the efficiency and efficacy of conservation endeavors. This research paper introduces an innovative study that centers on the creation and assessment of a novel deep learning (DL) model, which combines a Convolutional Neural Network (CNN) and Support Vector Machine (SVM). The model is specifically created to classify multiple wildlife species. By utilizing aerial imagery captured by drones, we have compiled a dataset consisting of 3397 photos with high levels of detail. These photographs encompass five unique categories, namely Mammals, Birds, Reptiles, Amphibians, and Fish. The evaluation of the model's performance was conducted based on metrics such as accuracy, precision, recall, and F1-score. Noteworthy accomplishments were demonstrated by an overall accuracy rate of 96.02%. The results of our study emphasize the potential of utilizing technology-driven methods to significantly transform wildlife conservation efforts. By automating the processes of species identification and monitoring, this approach can streamline data collection, provide valuable insights for conservation strategies, and contribute to the preservation of endangered species and their habitats. Our work highlights the significance of further research and ethical deliberations in the convergence of technology and conservation, given the persistent issues posed by visually similar species and variances within classes.