This is a ready-to-use python DenseCRF file for refining saliency maps using dense CRF, which is used in the NeurIPS 2021 paper "Joint Semantic Mining for Weakly Supervised RGB-D Salient Object Detection". If you want to know more about CRF, you can refer to here.
Test successfully on 2080Ti / RTX 3090 / Tesla P40 GPU
- Initial conda environment:
conda create -n JSM python=3.7. conda activate JSM- Install pydensecrf:
pip install git+https://github.com/lucasb-eyer/pydensecrf.git pip install --upgrade pip- Install opencv:
pip install opencv-python - Install numpy:
pip install numpy - Install PIL:
pip install Pillow - Install skimage:
pip install -U scikit-image
Check whether the installation is successful
cd DenseCRF/examples/python main.py- Check the outputs in
./outputfile.
- pydensecrf/densecrf/include/Eigen/Core:22:10: fatal error: ‘complex’ file not found
#include
^~~~~~~~~
1 warning and 1 error generated.
Command: conda install -c conda-forge pydensecrf