Implementation of "Consensus-Driven Propagation in Massive Unlabeled Data for Face Recognition" (CDP)

Notice

Modules for "Mediator" will be released soon.

Paper

Xiaohang Zhan, Ziwei Liu, Junjie Yan, Dahua Lin, Chen Change Loy, "Consensus-Driven Propagation in Massive Unlabeled Data for Face Recognition", ECCV 2018

Project Page: link

Dependency

Please use Python3, as we cannot guarantee its compatibility with python2. The version of PyTorch we use is 0.3.1. Other depencencies:

pip install nmslib

Before Start

Prepare your data list. If you want to evaluate the performance of CDP, copy the meta file as well. The example of list.txt and meta.txt can be found in data/example_data/.

mkdir data/your_data
cp /somewhere/list_file data/your_data/list.txt
cp /somewhere/meta_file data/your_data/meta.txt # optional

Prepare your feature files. Extract face features corresponding to the list.txt with your trained face models, and save it as binary files via feature.tofile("xxx.bin") in numpy. Finally link them to data/data_name/features/model_name.bin.

mkdir data/data_name/features
ln -s /somewhere/feature.bin data/your_data/features/resnet18.bin # for example

Although CDP can handle single-model case, we recommend more than one models to obtain better performance.

Prepare the config file. Please refer to the examples in experiments/

Usage

python -u main.py --config experiments/example_vote/config.yaml

python -u main.py --config experiments/example_mediator/config.yaml

Bibtex

@inproceedings{zhan2018consensus,
  title={Consensus-Driven Propagation in Massive Unlabeled Data for Face Recognition},
  author={Zhan, Xiaohang and Liu, Ziwei and Yan, Junjie and Lin, Dahua and Change Loy, Chen},
  booktitle={Proceedings of the European Conference on Computer Vision (ECCV)},
  pages={568--583},
  year={2018}
}

daaiyiyejian/cdp