Feature Pyramid Network on caffe

This is the unoffical version Feature Pyramid Network for Feature Pyramid Networks for Object Detection https://arxiv.org/abs/1612.03144

the mxnet unoffical version Feature Pyramid Network: https://github.com/unsky/Feature-Pyramid-Networks

results

FPN(resnet50) result is implemented without OHEM and train with pascal voc 2007 + 2012 test on 2007

[email protected]	aeroplane	bicycle	bird	boat	bottle	bus	car	cat	chair	cow

diningtable	dog	horse	motorbike	person	pottedplant	sheep	sofa	train	tv

framework

the red and yellow are shared params

about the anchor size setting

In the paper the anchor setting is Ratios： [0.5,1,2],scales :[8,]

With the setting and P2~P6, all anchor is [32,64,128,512,1024],but this setting is suit for COCO dataset which has so many small targets.

but the voc dataset has so many [128,256,512]targets.

So, we desgin the anchor setting:Ratios： [0.5,1,2],scales :[8,16], this is very import for voc dataset.

usage

download voc07,12 dataset ResNet50.caffemodel

• OneDrive download: link

compile caffe & lib

cd caffe-fpn
mkdir build
cd build
cmake ..
make -j16 all
cd lib
make 

train & test

./experiments/scripts/FP_Net_end2end.sh 1 FPN pascal_voc
./test.sh 1 FPN pascal_voc

feature pyramid networks for object detection

Lin, T. Y., Dollár, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. (2016). Feature pyramid networks for object detection. arXiv preprint arXiv:1612.03144.