Advanced FlexiNet-50 Model for Highly Accurate Multi-Class in Wildlife Monitoring Classification

Accurate animal classification in wildlife monitoring and conservation efforts is critical for understanding biodiversity and species behaviour. This study proposes the FlexiNet-50 model, a deep learning architecture, to improve classification accuracy across 15 animal classes. Leveraging the pre-trained ResNet50 model as a feature extractor, the architecture refines these features through fully connected layers with ReLU activation, BatchNormalization, and Dropout to enhance generalization and prevent overfitting. The FlexiNet-50 model achieved an accuracy of 97.7%, outperforming existing methods such as a hybrid deep residual CNN and a standard deep neural network framework. These results demonstrate the effectiveness of FlexiNet-50 in handling complex image classification tasks, particularly in diverse animal datasets. The model is highly accurate and has a strong foundation, making it useful for ecological research and animal welfare.