In the last few years, interest increased in addressing the concept of deep learning methods and investigating their performance on different machine learning and image processing tasks. With a similar experience, we have concentrated on a deep learning approach to reveal its performance on box-office prediction of movies. Due to its effectiveness, the Inception-V3 model has utilized to estimate box-offices by considering the prediction problem as a classification task. Across the experimental studies, the relation between movie poster and box-office of movies have been explored and the prediction performance of system has noted with %33 accuracy rate from conducting experiments on 1307 images.