Gangs are a serious threat to the public safety in the United States. We have developed a system known as Gang Graffiti Automatic Recognition and Interpretation (GARI) to help law enforcement identify, track, and analyze gang activities. Gang graffiti components are the segmented graffiti content including symbols, digits, and characters. In this paper, we propose a deep convolutional neural network to classify the graffiti components. We make a comparison between our proposed deep learning method and our previous traditional method. Experimental results show the proposed method reaches 89.3% accuracy with dropout regularization.