Dialect Recognition Systems (DRS) are systems that group dialects, according to similar acoustic features found in dialect regions. The speaker's age, gender, and dialect characteristics negatively affect the performance of speech recognition systems. To handle dialect differences, dialect recognition systems can be integrated into speech recognition systems. By determining the spoken dialect, the system can be switched to the corresponding speech recognition model. There is no dataset that can be used for Turkish automatic dialect recognition systems. In this study, it is thought that this deficiency should be eliminated in some way. In addition, an experimental study has been carried out to classify the generated data set by convolutional neural networks. The resulting 83.3% accuracy is satisfactory.