Dynamic Feature Pyramid Networks for Object Detection. arXiv
By Mingjian Zhu, Kai Han, Changbin Yu, Yunhe Wang
This is the implementation of DyFPN. Basically, we follow the setting of testing a model in MMDetection. Please refer to MMDetection for installation and dataset preparation.
-
Download the pre-trained model in Onedrive.
-
Create a new folder named checkpoint and put the pre-trained model in it.
-
Test the model with the following command.
python tools/test.py configs/dyfpn/faster_rcnn_r50_dyfpn_1x_coco.py checkpoints/DyFPN_B_CNNGate.pth --eval bbox
@article{zhu2020dynamic,
title={Dynamic Feature Pyramid Networks for Object Detection},
author={Zhu, Mingjian and Han, Kai and Yu, Changbin and Wang, Yunhe},
journal={arXiv preprint arXiv:2012.00779},
year={2020}