DetectionTeamUCAS / NAS_FPN_Tensorflow

NAS-FPN: Learning Scalable Feature Pyramid Architecture for Object Detection.
https://arxiv.org/abs/1904.07392
MIT License
214 stars 62 forks source link

Question on top-down layer in NAS-FPN #13

Open ggjy opened 5 years ago

ggjy commented 5 years ago

Hi, it seems that there is no top-down path for NAS-FPN (https://github.com/tensorflow/tpu/blob/master/models/official/detection/modeling/architecture/nasfpn.py) in the official implement? I am not sure about this.

这里的实现是把NAS-FPN接在top-down之后的feature map上么?官方代码好像是没有top-down那步?我自己在pytorch上的实现发现没有top-down的话NAS-FPN会掉点,不知道是不是我哪理解错了?