In the field of deep learning, transfer learning has emerged as a key strategy, particularly when applying pre-trained visual models to improve various computer vision tasks. With a focus on using a range of pre-trained convolutional neural network (CNN) architectures for image categorization, comprising VGG16, DenseNet, MobileNet, Inception v3, ResNet-50, and Xception, this study investigates the use of transfer learning in image classification. By conducting a comprehensive evaluation of these models concerning both training and testing accuracy, valuable insights are offered on their relative effectiveness. DenseNet121 and VGG16 are top performers, and these two algorithms demonstrate good efficiency, particularly in situations where data availability is limited. Furthermore, this study expands its scope to include the classification of fast radio bursts (FRBs), showcasing the effectiveness of transfer learning, particularly with the VGG16 and DenseNet121. The results indicate that transfer learning with pre-trained CNN models holds the potential for enhancing data analysis in astronomical studies, consequently deepening our comprehension of cosmic phenomena like FRBs.