@inproceedings{connectingvisionandlanguagewithvideo,
  title = {Connecting Vision and Language with Video Localized Narratives},
  author = {Paul Voigtlaender and Soravit Changpinyo and Jordi Pont-Tuset and Radu Soricut and Vittorio Ferrari},
  year = {2023},
  booktitle = {CVPR 2023 1},
  eprint = {2302.11217},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2302.11217v2},
}