One-shot Learning for Semantic Segmentation (OSLSM)

By Amirreza Shaban, Shray Bansal, Zhen Liu, Irfan Essa and Byron Boots

Paper

You can find our paper at https://arxiv.org/abs/1709.03410

Citation

If you find OSLSM useful in your research, please consider to cite:

@inproceedings{shaban2017one,
 title={One-Shot Learning for Semantic Segmentation},
 author={Shaban, Amirreza and Bansal, Shray and Liu, Zhen and Essa, Irfan and Boots, Byron},
 journal={British Machine Vision Conference ({BMVC})},
 year={2017}
}

Instructions for Testing (tested on Ubuntu 16.04)

We assume you have downloaded the repository into ${OSLSM_HOME} path.

Install Caffe prerequisites and build the Caffe code (with PyCaffe). See http://caffe.berkeleyvision.org/installation.html for more details

cd ${OSLSM_HOME}
mkdir build
cd build
cmake ..
make all -j8

If you prefer Make, set BLAS to your desired one in Makefile.config. Then run:

cd ${OSLSM_HOME}
make all -j8
make pycaffe

Update the $PYTHONPATH:

export PYTHONPATH=${OSLSM_HOME}/OSLSM/code:${OSLSM_HOME}/python:$PYTHONPATH

Download PASCAL VOC dataset: http://host.robots.ox.ac.uk/pascal/VOC/voc2012/
Download trained models from: https://gtvault-my.sharepoint.com/:u:/g/personal/ashaban6_gatech_edu/EXS5Cj8nrL9CnIJjv5YkhEgBQt9WAcIabDQv22AERZEeUQ
Set CAFFE_PATH=${OSLSM_HOME} and PASCAL_PATH in ${OSLSM_HOME}/OSLSM/code/db_path.py file
Run the following to test the models in one-shot setting:

cd ${OSLSM_HOME}/OSLSM/os_semantic_segmentation
python test.py deploy_1shot.prototxt ${TRAINED_MODEL} ${RESULTS_PATH} 1000 fold${FOLD_ID}_1shot_test

Where ${FOLD_ID} can be 0,1,2, or 3 and ${TRAIN_MODEL} is the path to the trained caffe model. Please note that we have included different caffe models for each ${FOLD_ID}.

Simillarly, run the following to test the models in 5-shot setting:

cd ${OSLSM_HOME}/OSLSM/os_semantic_segmentation
python test.py deploy_5shot.prototxt ${TRAINED_MODEL} ${RESULTS_PATH} 1000 fold${FOLD_ID}_5shot_test

For training your own models, we have included all prototxts in ${OSLSM_HOME}/OSLSM/os_semantic_segmentation/training directory and the vgg pre-trained model can be found in snapshots/os_pretrained.caffemodel.

You will also need to

Download/Prepare SBD dataset (http://home.bharathh.info/pubs/codes/SBD/download.html).
Set SBD_PATH in ${OSLSM_HOME}/OSLSM/code/db_path.py
Set the profile to fold${FOLD_ID}\_train for our data layer (check the prototxt files and ${OSLSM_HOME}/OSLSM/code/ss_datalayer.py) to work.

License

The code and models here are available under the same license as Caffe (BSD-2) and the Caffe-bundled models (that is, unrestricted use; see the BVLC model license).

Contact

For further questions, you can leave them as issues in the repository, or contact the authors directly:

Amirreza Shaban amirreza@gatech.edu

Shray Bansal sbansal34@gatech.edu

Zhen Liu liuzhen1994@gatech.edu

About Caffe

Caffe is a deep learning framework made with expression, speed, and modularity in mind. It is developed by Berkeley AI Research (BAIR)/The Berkeley Vision and Learning Center (BVLC) and community contributors.

Check out the project site for all the details like

and step-by-step examples.

Custom distributions

Intel Caffe (Optimized for CPU and support for multi-node), in particular Xeon processors (HSW, BDW, SKX, Xeon Phi).
OpenCL Caffe e.g. for AMD or Intel devices.
Windows Caffe

Community

Please join the caffe-users group or gitter chat to ask questions and talk about methods and models. Framework development discussions and thorough bug reports are collected on Issues.

Happy brewing!

License and Citation

Caffe is released under the BSD 2-Clause license. The BAIR/BVLC reference models are released for unrestricted use.

Please cite Caffe in your publications if it helps your research:

@article{jia2014caffe,
  Author = {Jia, Yangqing and Shelhamer, Evan and Donahue, Jeff and Karayev, Sergey and Long, Jonathan and Girshick, Ross and Guadarrama, Sergio and Darrell, Trevor},
  Journal = {arXiv preprint arXiv:1408.5093},
  Title = {Caffe: Convolutional Architecture for Fast Feature Embedding},
  Year = {2014}
}