This work consists of systematically reviewing studies involving applications of convolutional neural network (CNN) algorithms implemented in single board computers (SBC) to verify their feasibility in computer vision (CV) applications. Indicating the main parameters for evaluating the performance of SBCs as the processing time (88.24%) and accuracy (29.41%). Furthermore, the most used SBCs and the motivation for choosing such devices are indicated. The study contributes to help in future works involving CNNs in embedded systems, as well as in discussions about architectures involving deep learning applications.