The reference points will be updated during each layer of decoder?

fundamentalvision / Deformable-DETR

Deformable DETR: Deformable Transformers for End-to-End Object Detection.

Apache License 2.0

3.22k stars 520 forks source link

Open GivanTsai opened 2 years ago

GivanTsai commented 2 years ago

The reference points are generated through a linear projection of object queries. Are they updated during the iteration of 6 decoder layers?