Wheat is one of the most produced and consumed grain products worldwide. Wheat is the main grain product in developed and underdeveloped countries. Flour obtained from wheat is used in the production of bread, the most basic food product, and in the production of cakes used to celebrate the most special days. Therefore, knowing the pure bread wheat varieties is important both for production and for those who use wheat as flour. However, since wheat varieties are very similar to each other, it is difficult to distinguish them. To solve this problem, a pre-trained hybrid model based on convolutional neural network (CNN) is proposed in this study to classify bread wheat varieties. Images of five different registered bread wheat varieties were captured and a bread wheat image data set was created by separating them with image processing techniques to be used in deep learning. Then, the obtained images were classified using transfer learning with fine-tuning on the Xception model, one of the pre-trained CNN models. To increase the classification success, Xception CNN model and BiLSTM (Bidirectional Long Short-Term Memory) algorithms hybrid (Xception + BiLSTM) models were obtained. As a result of classifications, the highest classification success was obtained from the Xception + BiLSTM model with 97.73%. The results revealed that the proposed methods can be used in systems used for classification of bread wheat varieties and to obtain pure wheat varieties automatically.