SwinBERT: End-to-End Transformers with Sparse Attention for Video Captioning | IEEE Conference Publication | IEEE Xplore