This study was conducted to automatically detect whether signature images are genuine or forged using the Darknet19 model and feature selectors. Initially, various features were extracted from the signature images. Feature selectors such as mRMR, Chi2, ReliefF, ANOVA, and Kruskal Wallis were employed to choose these features. According to the results of the study, the accuracy results obtained without feature selection are as follows: No feature selector yields 97.92%, mRMR yields 98.83%, Chi2 yields 98.48%, ANOVA yields 98.56%, Kruskal Wallis yields 98.60%, and ReliefF feature selector achieves an accuracy of 99.32%. The results indicate that the ReliefF feature selector provides the highest accuracy. The results obtained using the ReliefF feature selector showed a 1.4% increase in accuracy compared to the results obtained without feature selection. This study demonstrates that artificial intelligence and feature selection methods can effectively detect signature forgery.