Implementation of MobileNet V3 In Classifying Butterfly Species with Android and Cloud Based Application Development

Ihsan Zulfahmi
Said Iskandar Al Idrus
Hermawan Syahputra
Insan Taufik
Kana Saputra S

Journal of Artificial Intelligence and Engineering Applications (JAIEA)
Vol. 4 No. 2 (2025): February 2025

This research aimed to develop an Android application capable of classifying butterfly species using cloud computing and deep learning technologies. MobileNetV3-Large, a Convolutional Neural Network (CNN) architecture, was employed to process and classify six butterfly species. The dataset was divided into two ratios, 70:30 and 80:20, for training and testing. Evaluation results indicated that the optimal model was achieved with an 80:20 ratio, yielding an accuracy of 94% and precision, recall, and F1-Score values exceeding 90% for each species class. Google Cloud Platform (GCP) was utilized to manage and run the model using the Cloud Run service, enabling the application to function efficiently even with limited resources on Android devices. The application incorporates an encyclopedia of species and a camera scanning feature, making it a valuable educational tool