search

PaddlePaddle / models

Officially maintained, supported by PaddlePaddle, including CV, NLP, Speech, Rec, TS, big models and so on.
Apache License 2.0
6.9k stars 2.91k forks source link

pyramidbox: why anchor down-sampled? #1860

Open chenkingwen opened 5 years ago

chenkingwen commented 5 years ago

I have a question disturbed me. As we know, the anchor will be expanded to detect head and body from face anchors, but it is down-sampled to calculate IOU with region target to confirm the label. After all, the head and body are larger than face. Thank you very much.

takecareofbigboss commented 5 years ago

hi, maybe it is the same when they are mapping the the original image.

takecareofbigboss commented 5 years ago

@chenkingwen

chenkingwen commented 5 years ago

@takecareofbigboss I do not understand the image pyramid mean what in paper. The semi-supervised anchor would be used in upper two feature map from current layer? but why down-sampled? some one help me? thanks!

takecareofbigboss commented 5 years ago

maybe you can see figure 1 for reference~