AmigoCDT/iqiyi-vid-challenge

Code for IQIYI-VID Challenge based on ESSH and Insightface

Recently www.iqiyi.com released a great video person dataset called IQIYI_VID and also launched a person search competition on it. It is a very large and real dataset worth trying to verify your face model accuracy precisely.

This repository contains the code for IQIYI-VID(IQIYI video person identification) Challenge. The methods are implemented in Python and MXNet. The Enhanced SSH (ESSH) from enhanced-ssh-mxnet is applied for face detection and alignment. Insightface scheme is used for face recognition.

Pre-trained models can be downloaded on BaiduCloud or GoogleDrive.

Environment

This repository has been tested under the following environment:

Python 2.7
Ubuntu 18.04
Mxnet-cu90 (==1.3.0)

Installation

Prepare the environment.
Clone the repository.
Type make to build necessary cxx libs.
Download the pre-trained model and place it in ./model/
Download the IQIYI-VID Datasets from IQIYI_VID and unzip them to your disk.

Usage

Use python detect.py to train+val dataset and test dataset respectively for face detection.
Use python feature.py to detections of train+val and test dataset respectively for feature extraction.
Use python genfeat.py to re-save the extracted face features for training the MLP network.
Run train_mlp.py to train the MLP network for face ID recognition using train+val datasets.
Run python predict.py to features of test dataset for predicting face ID using the trained MLP network.
Run python submit.py to output the final submissions for IQIYI-VID Challenge.

License

MIT LICENSE

Reference

@article{deng2018arcface,
title={ArcFace: Additive Angular Margin Loss for Deep Face Recognition},
author={Deng, Jiankang and Guo, Jia and Niannan, Xue and Zafeiriou, Stefanos},
journal={arXiv:1801.07698},
year={2018}
}

@inproceedings{Najibi2017SSH,
  title={SSH: Single Stage Headless Face Detector},
  author={Najibi, Mahyar and Samangouei, Pouya and Chellappa, Rama and Davis, Larry S.},
  booktitle={IEEE International Conference on Computer Vision},
  year={2017},
}

Acknowledgment

The code is adapted based on an intial fork from the insightface repository.

AmigoCDT / iqiyi-vid-challenge