search

fundamentalvision / Deformable-DETR

Deformable DETR: Deformable Transformers for End-to-End Object Detection.
Apache License 2.0
3.15k stars 513 forks source link

DeformableTransformerDecoder, always sets self.bbox_embed = None? #177

Open vadimkantorov opened 1 year ago

vadimkantorov commented 1 year ago

Hi! It appears that DeformableTransformerDecoder.init always sets self.bbox_embed = None, so if self.bbox_embed is not None in forward(...) is never triggered.

Why is that? Am I missing anything?

Thanks!

https://github.com/fundamentalvision/Deformable-DETR/blob/main/models/deformable_transformer.py#L321-L324:

# in DeformableTransformerDecoder.__init__:

# hack implementation for iterative bounding box refinement and two-stage Deformable DETR
self.bbox_embed = None
self.class_embed = None