BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition

Official PyTorch implementation of BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition (https://arxiv.org/abs/2008.06674), ECCV 2020

This repository contains source codes of experiments for image retrieval (deep metric learning).

Dependencies

You need a CUDA-enabled GPU and python (>3.6) to run the source code.

• torchvision >= 0.4.2
• torch >= 1.1.0
• tqdm
• scipy
• Pillow

pip install -r requirements.txt

Preparing datasets

1. Make dataset directory

mkdir ./dataset

2. (Optional) Only for In-Shop Clothes Retrieval

The source code will automatically download CUB-200-2011, Cars-196, and Stanford Online Products datasets.

But you need to manually download In-Shop Clothes Retrieval dataset.

1. Make Inshop directory in ./dataset directory

mkdir -p ./dataset/Inshop

2. Download img.zip at the following link, and unzip it in Inshop directory

https://drive.google.com/drive/folders/0B7EVK8r0v71pYkd5TzBiclMzR00

3. Download list_eval_partition.txt at the following link, and put it in the Inshop directory.

https://drive.google.com/drive/folders/0B7EVK8r0v71pWVBJelFmMW5EWnM

Quick start

# Stanford Online Products
### ArcFace
python run_stanford.py --loss arcface

### BroadFace (without Compensation)
python run_stanford.py --loss broadface --queue-size 32000

### BroadFace (with Compensation)
python run_stanford.py --loss broadface --queue-size 32000 --compensate


# In-Shop Clothes Retrieval
### ArcFace
python run_inshop.py --loss arcface

### BroadFace (without Compensation)
python run_inshop.py --loss broadface --queue-size 32000

### BroadFace (with Compensation)
python run_inshop.py --loss broadface --queue-size 32000 --compensate

Citation

In case of using this source code for your research, please cite our paper.

@inproceedings{kim2020broadface,
  title={BroadFace: Looking at Tens of Thousands of People at Once for Face Recognition},
  author={Kim, Yonghyun and Park, Wonpyo and Shin, Jongju},
  booktitle={European Conference on Computer Vision},
  pages={536--552},
  year={2020},
  organization={Springer}
}

Contact

Yonghyun Kim

Wonpyo Park

Jongju Shin

Kakao Enterprise/Vision Team