Hybrid Fusion Approach for Animal Species Identification using Facial Images

Species identification is crucial in the conservation of biodiversity, assessment, and monitoring of wildlife and features of an ecosystem. Disadvantages of conventional techniques include low accuracy, variance of environmental factors, and unbalanced data set. It discusses a novel hybrid species identification approach using the ResNet-50 CNN. To improve the quality of inputs, the methodology uses image preprocessing, which encompasses filtering, resizing, and data augmentation. The performance of the model was measured while training and testing it on a diverse group of species captured in various climates. The classifiers for the proposed framework offered a classification accuracy of 94.8%, an average inference per image time of 48 milliseconds, and better precision and recall comparisons to past models. These results highlight the framework's application for real-time wildlife tracking and monitoring. It provides a relevant contribution to environmental issues and provides a practical approach to solving ecological issues.