This code is for the paper "DRNet: Double Recalibration Network for Few-Shot Semantic Segmentation". The implementation of the DRNet* is in drnetv2
- Python 2.7
- Pytorch 1.3.1
The dependencies can be installed by running:
pip install -r requirements.txt
We provide trained Models with the ResNet-50 backbone on Pascal-5i and COCO-20i for performance evalution. You can download from here.
- Change the dataset path in db_path.py, ss_datalayer.py line 342 and train_frame_coco.py.
- Pascal-5i
- Run the train scripts
cd scripts ./train_groupX.sh
- Run the evalution scripts for Pascal-5i
python test_frame_all.py
- Run the train scripts
- COCO-20i
- Run the train scripts
cd scripts ./coco_train_groupX.sh
- Run the evalution scripts for Pascal-5i
python test_frame_all_coco.py
- Run the train scripts