Enhanced Classification of Intel Images Using Refined EfficientNet and MobileNetV2 Frameworks

The proposed study describes the functioning of two modern deep learning models, namely EfficientNet and MobileNetV2, within the scope of image classification from a wide range of datasets. This study is designed to categorize images into six groups, namely: buildings, mountains, forests, sea/ocean, glaciers, and streets. The study utilizes a dataset from the Kaggle repository to assess the effectiveness of these models. The research methodology will include data preprocessing, modeling, and model evaluation. This paper will find out the balance between accuracy and efficiency which is provided by each model. EfficientNet is compared with MobileNetV2 because of its scalable structure and compromise between model depth, width, and resolution versus the latter which is known for its low weight and efficiency, especially in constrained circumstances. The research utilizes a framework of accuracy, precision, recall, F1-score metrics, and model efficiency analyses for comprehensive result evaluation and scrutiny. The analysis of this study indicates the unique advantages of separate models, for instance, EfficientNet is more accurate in most cases and MobileNetV2 is a good solution when you have to save computer resources. Comparing models highlights the importance of the selection that is made based on the specific application instead of performance metrics or computational demands.