Should I train a new backbone?

  Kiến thức lập trình

I would like to make a semantic segmentation model based on Key-Value Transformer.. In this paper, query has removed from self-attention part.

Therefore, I am not sure if I could still use ViT as my backbone. If not possible, should I make a new backbone or is there any other solution?

  • pytorch
  • semantic-segmentation
  • pre-trained-model
  • vision-transformer

LEAVE A COMMENT