A web-based application to classify cendrawasih birds using deep learning

Abstract

This research aims to develop a deep learning-based model to classify images of birds of paradise
(Cendrawasih). Three different model architectures were employed in this study: Convolutional Neural Network (CNN),
InceptionResNetV2, and MobileNetV2. The dataset consists of several species of birds of paradise, which were
processed using data augmentation techniques to enhance the variety and quality of the training data. The model training
and evaluation processes were conducted using TensorFlow and Keras, with the application of callbacks such as
EarlyStopping to prevent overfitting. Evaluation results indicate that the MobileNetV2 and InceptionResNetV2 models
achieved the highest accuracy, with an average score above 90%. The system implementation involved developing a web
application based on Flask and React JS to facilitate real-time image prediction. This study demonstrates the
effectiveness of using deep learning models for bird of paradise image classification and their potential for application in
limited computing environments.