Similar Items: Automatic video captioning using spatiotemporal convolutions on temporally sampled frames