These are the trained models for our adaptation of
Pix2Seq for token-based autoregressive video object detection and semantic segmentation.
We also include all the
pretrained models we have used in our experiments as well as a
sample of evaluation data.
