Transfer learning is recognized as a widely utilized and crucial technology that is used to facilitate the transfer of knowledge from one domain to solve new problems. Various classification tasks, including flower classification, are extensively addressed using transfer learning, which introduces unique challenges because of wide diversity of flower species, many of which share similar shapes, appearances, or environmental features like leaves and grass. In this paper, we introduce AMobileNet, an enhanced version of traditional MobileNetV2 model, specifically optimized for flower classification. AMobileNet uses the Adam optimizer and SparseCategoricalCrossEntropy loss function to perform better on the Kaggle flower dataset. The Adam optimizer, also known as Adaptive Moment Estimation, is one of the most popular optimization algorithms widely used for training machine learning and deep learning models. This has been designed for the purpose of combining two other popular optimization algorithms: AdaGrad and RMSProp. Adam itself is relatively memory-efficient, only requiring a small amount of memory. As well, it is apt for very large datasets or high-dimensional parameter spaces. Furthermore, the SparseCategoricalCrossEntropy loss function is an efficient choice for multiclass classification tasks. We have assessed the classification performance of the AMobileNet model and compared it with the traditional CNN architecture. It has been observed from the findings that AMobileNet performs better than CNN, ensuring the effectiveness in distinguishing between various flower pattern varieties. Therefore, this strategy may be quite useful in a number of applications: botany, agriculture, and so forth, where flower classification is an important issue.