Although new coating development for improved surface protection is necessary, its manual application has been a difficult problem to solve. In this study, a convolution neural network (CNN) was used for prediction of the painting gun operation. Painting videos were converted to sequential images, of which two consecutive images were associated with the gun position in the next time step. The inputs were implemented in a regression CNN training, which was used to calculate the position of the spray gun at the next moment. Recursive image utilization provides the prediction of spray gun movement in real-time applications. The statistical measures of the prediction and true values of gun movement using test data indicate that the proposed CNN gives comparable outcomes to similar applications of the CNN. The exhibition of simulated painting of a rectangle and a semicircle demonstrates the usefulness of the proposed CNN application for spray gun painting.