py-R-FCN

R-FCN

Description

R-FCN is a type of region-based object detector. The whole network is mainly composed of full convolution network ( ResNet ) and RPN network. The former is used to extract features, and the latter is used to generate ROI.

Model

Model	Download	mAP
R-FCN	model weight	27.9

Dataset

• COCO 2014 train.
• COCO 2014 val.

References

• R-FCN: Object Detection via Region-based Fully Convolutional Networks
• YuwenXiong/py-R-FCN

License

MIT