search

megvii-research / AnchorDETR

An official implementation of the Anchor DETR.
Other
338 stars 36 forks source link

Hello! The Feature position in figure3 is the positional encoding in DETR #36

Closed cxq1 closed 2 years ago

cxq1 commented 2 years ago

Hello! The Feature position in figure3 is the positional encoding in DETR

tangjiuqi097 commented 2 years ago

No, the feature position is the normalized grid position of the image feature map. For example, if the shape (H, W) of feature map is 4x5, then the feature position is: image

You can refer to the code for more detail.

github-actions[bot] commented 2 years ago

This issue is not active for a long time and it will be closed in 5 days. Feel free to re-open it if you have further concerns.