Discriminative Visual Similarity Search with Semantically Cycle-consistent Hashing Networks

Here is the official codebase for Discriminative Visual Similarity Search with Semantically Cycle-consistent Hashing Networks. MMAsia2022 Version
Extended TOMM Version
This repo contains two versions of SCHN(the original SCHN on MMAsia2022 and the one with the discriminator TOMM). You can specify the variant by the method_name constant in run.py(Line 18).

REQUIREMENTS

requirements.txt contains libraries used in my environments. Though other versions may also work, I have no time to test and can't guarantee any of that.

DATASETS

1. CIFAR-10

2. NUS-WIDE Password: uhr3

3. MIR-Flickr 25K version

Citation:

If you find our approach useful in your research, please cite our papers:
[1] Z. Zhang, J. Wang, G. Lu, Towards Discriminative Visual Search via Semantically Cycle-consistent Hashing Networks, in Proc. of The ACM Multimedia Asia, 2021.
[2] Z. Zhang, J. Wang, L. Zhu, G. Lu, Discriminative Visual Similarity Search with Semantically Cycle-consistent Hashing Networks, ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM), 18(2s): 1-21, 2022.

@incollection{zhang2021towards,
  title={Towards Discriminative Visual Search via Semantically Cycle-consistent Hashing Networks},
  author={Zhang, Zheng and Wang, Jianning and Lu, Guangming},
  booktitle={ACM Multimedia Asia},
  pages={1--7},
  year={2021}
}

@article{zhang2022discriminative,
  title={Discriminative Visual Similarity Search with Semantically Cycle-consistent Hashing Networks},
  author={Zhang, Zheng and Wang, Jianning and Zhu, Lei and Lu, Guangming},
  journal={ACM Transactions on Multimedia Computing, Communications, and Applications (TOMM)},
  volume={18},  
  number={2s},  
  pages={1--21},  
  year={2022},
  publisher={ACM New York, NY}
}

Project Structure

The structure of the project files should go as follows.

.
├── data
│   ├── cifar10.py
│   ├── data_loader.py
│   ├── flickr25k.py
│   ├── imagenet.py
│   ├── __init__.py
│   ├── nus_wide.py
│   └── transform.py
├── dataset
│   ├── cifar-10-batches-py
│   │   ├── batches.meta
│   │   ├── data_batch_1
│   │   ├── data_batch_2
│   │   ├── data_batch_3
│   │   ├── data_batch_4
│   │   ├── data_batch_5
│   │   ├── readme.html
│   │   └── test_batch
│   ├── Flickr
│   │   ├── database_label.txt
│   │   ├── db_image.txt
│   │   ├── mirflickr
│   │   │   ├── im21110.jpg
│   │   │   └── ......
│   │   ├── test_image_m.txt
│   │   └── test_label.txt
│   └── NUS-WIDE
│       ├── database_img.txt
│       ├── database_label_onehot.txt
│       ├── database_label.txt
│       ├── images
│       │   ├── 0068_2569963337.jpg
│       │   └── ......
│       ├── img_tc10.txt
│       ├── README.md
│       ├── targets_onehot_tc10.txt
│       ├── targets_tc10.txt
│       ├── test_img.txt
│       ├── test_label_onehot.txt
│       ├── test_label.txt
│       └── t.txt
├── logs
├── main_mmasia.py
├── main_tomm.py
├── modules
│   ├── alexnet.py
│   ├── all_in_one.py
│   ├── gcn
│   │   ├── layers.py
│   │   ├── models.py
│   │   └── __pycache__
│   │       ├── layers.cpython-39.pyc
│   │       └── models.cpython-39.pyc
│   ├── __init__.py
│   └── mlp.py
├── README.md
├── requirements.txt
├── run.py
└── utils
    ├── evaluate.py
    └── __init__.py