IEMask R-CNN: Information-enhanced Mask R-CNN

by Xiuli Bi, Jinwu Hu, Bin Xiao*, Weisheng Li, Xinbo Gao

This is an official implementation of IEMask R-CNN in our IEEE Transactions on Big Data paper " IEMask R-CNN: Information-enhanced Mask R-CNN"

Usage

Note

Our code is based on the Detectron2 and BMask R-CNN implementation.

Training

python tools/train_net.py --config-file configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml --num-gpus 1

Evaluation

specify a config file and test with trained model

python train_net.py --config-file configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_1x.yaml --num-gpus 1 --eval-only MODEL.WEIGHTS ./output/model_final.pth

Citation

If you use IEMask R-CNN in your research, please cite our IEEE Transactions on Big Data paper.

@ARTICLE{9811396,
  author={Bi, Xiuli and Hu, Jinwu and Xiao, Bin and Li, Weisheng and Gao, Xinbo},
  journal={IEEE Transactions on Big Data}, 
  title={IEMask R-CNN: Information-Enhanced Mask R-CNN}, 
  year={2023},
  volume={9},
  number={2},
  pages={688-700},
  doi={10.1109/TBDATA.2022.3187413}}

update status

The code (V1) is uploaded (Ongoing updates).