An End-to-End CNN-Based Framework for Butterfly Species Recognition and Visualization via Web Interface

Abstract

Butterflies are vital indicators of environmental health, and accurate species identification is essential for ecological research and conservation. This paper presents a deep learning-based butterfly classification system integrated with a web interface for real-time
image-based prediction. The system utilizes a Convolutional Neural Network (CNN) to classify 75 butterfly species using labeled image data. To improve generalization, the model employs data augmentation techniques such as rotation, zooming, and flipping. The CNN architecture consists of multiple convolutional and pooling layers followed by dense layers, enabling effective feature extraction
and classification. The model is trained over 40 epochs using the Adam optimizer and categorical cross-entropy loss, achieving high training and validation accuracy. A user-friendly Django web application is developed to allow users to register, log in, upload butterfly
images, and receive real-time predictions. The system displays the predicted species along with the uploaded image, providing an accessible platform for students, researchers, and enthusiasts. By combining deep learning with a web-based interface, this work offers a practical and interactive solution for automated butterfly species identification, demonstrating the effectiveness of CNNs in ecological image classification tasks.