jinglescode/papers

End-to-End Video Instance Segmentation with Transformers

jinglescode opened this issue · 1 comments

jinglescode commented

Paper

Link: https://arxiv.org/pdf/2011.14503.pdf
Year: 2020

Summary

  • end-to-end instance segmentation on video frames, tracking the object across frames
  • achieves the highest speed among all existing video instance segmentation models, and achieves the best result

image

Contributions and Distinctions from Previous Works

  • simplifying the overall pipeline and is significantly different from existing approaches

Methods

image

mtachaudhary commented

Hi @jinglescode
I hope you will be doing well.

Can you please share the source code of this paper? I shall be very thankful to you for this favor.

Thank you