Hybrid Ensemble of MobileNetV2 and EfficientNetB0 for Enhanced Land Use and Land Cover Classification
Likhitesh Pekala, Gunnam Suryanarayana, Siva Kumar Kuttuboina, Moksha Yagna Kumar Anagani · 2025
In this research, we instigated a composite neural network approach by integrating MobileNetV2 and EfficientNetB0 models to boost the precision and efficiency of land use and land cover (LULC) classification. Utilizing the NWPU-RESISC45 dataset and SIRI-WHU, which encompasses a diverse set of remote-sensing images across 12 distinct classes, our model leverages MobileNetV2’s efficient architecture and EfficientNet-B0’s high accuracy. The combined model achieved a notable 98.3% and 96.04% accuracy respectively for the datasets, affirming its effectiveness and robustness in categorizing various types of land cover in complex scenes. This performance highlights the model’s potential for scalable, real-time environmental monitoring and urban planning.