Bone fractures are a prevalent medical problem that requires prompt intervention to prevent enduring effects. Although conventional diagnostic techniques such as X-rays are prevalent, they require skilled radiologists to interpret the precise image. Recent advances in deep learning, especially Convolutional Neural Networks (CNNs), have demonstrated significant potential to automate and improve the precision of fracture identification. In this study, the application of robust CNN-based deep learning methods is presented for the identification of fractures. We investigated the utilization of CNNs on a large data set (FracAtlas) of medical images, illustrating the model’s ability to classify the existence of fractures autonomously. Our model achieves a substantial accuracy of 98. 7% and a sensitivity of 100%, facilitating an expedited diagnosis and aiding physicians in their decision-making process compared to a class attention transformer, which only achieves 53%. Furthermore, we compare conventional approaches and cutting-edge deep learning models, emphasizing CNN’s enhanced efficacy in recognizing bone fractures.