Among the most frequent injuries that physicians encounter in routine clinical settings are bone fractures. To find these fractures, radiologists typically examine X-ray pictures, although this procedure can be laborious and frequently differs among specialists. Our paper presents a deep learning-based method for classifying bone fractures as present or absent in order to simplify and enhance this task. We constructed a comprehensive pipeline that consists of a popular transfer learning model, MobileNetV2, as well as a specially created CNN. A publicly accessible dataset from Kaggle, comprising 4,097 training imageries, 404 for validation, and 399 for testing, was used to train and evaluate these models. We used a variety of data augmentation and regularization strategies to make sure the models generalize effectively. With a test accuracy of 100%, sensitivity of 100%, and an AUC of 100%, MobileNetV2 outperformed the other model. Our results demonstrate the ability of deep convolutional neural networks to reliably identify bone fractures, which could provide important assistance in clinical diagnosis.