Tie-Aware Hashing

This repository contains Matlab/MatConvNet implementation for the following paper:

"Hashing as Tie-Aware Learning to Rank",
Kun He, Fatih Cakir, Sarah Adel Bargal, and Stan Sclaroff. IEEE CVPR, 2018 (arXiv)

If you use this code in your research, please cite:

@inproceedings{He_2018_TALR,
  title={Hashing as Tie-Aware Learning to Rank},
  author={Kun He and Fatih Cakir and Sarah Adel Bargal and Stan Sclaroff},
  booktitle={The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  month={June}, 
  year={2018}
}

Preparation

• Create or symlink a directory cachedir under the root directory to hold experimental results
• Install or symlink MatConvNet at ./matconvnet (for training CNNs)
• Install or symlink VLFeat at ./vlfeat. Note: this is only necessary for computing the regular tie-agnostic AP metric. We provide efficient implementation for the tie-aware metrics in +eval.
• Download datasets and pretrained models: see data/README.md.

Usage

• In the root folder, run startup.m
• To (approximately) replicate results in the paper, run one of the run_*.m files. For example, run_cifar_s1(32) will run the Setting 1 experiment on the CIFAR-10 dataset, with 32-bit hash codes, using the default parameters therein.
• Alternatively, directly run the scripts in +demo/ with your parameter choices. See main/get_opts.m for the parameters.

License

MIT License, see LICENSE

Contact

For questions/comments, feel free to contact:

kunhe@fb.com

Notes

• We provide simplified versions of tie-aware AP (apr_s_forward.m and apr_s_backward.m) and tie-aware NDCG (ndcgr_s_forward.m and ndcgr_s_backward.m). They attain similar performance compared to the original versions, but are much simpler to implement. The derivations can be found in the appendix of the arxiv version of the paper (v4).
• This implementation is partly based on MIHash.