Convolutional Neural Networks (CNNs) have demonstrated remarkable performance in various applications, particularly in the field of medical imaging. A notable application gaining attention is bone age prediction, which involves predicting bone development based on radiography images. In this project, we implemented and evaluated multiple CNN-based architectures, namely MobileNet, Inception V3, and Xception. The crucial aspect of any deep learning project lies in the selection of hyperparameters. To address this, we employed Keras Tuner for optimal parameter selection. Our best-performing model, Xception 1, achieved a mean absolute error (MAE) of 9.79, followed by Inception V3 with 12.04, and Xception 2 with 12.48