lucasxlu / crnet Goto Github PK

View Code? Open in Web Editor NEW

Official PyTorch Implementation of Paper <CRNet: Classification and Regression Neural Network for Facial Beauty Prediction> (Pacific-Rim Conference on Multimedia (PCM) 2018)

License: MIT License

Python 100.00%

crnet's Introduction

CRNet: Classification and Regression Neural Network for Facial Beauty Prediction

Introduction

This repository holds the PyTorch implementation for our paper CRNet: Classification and Regression Neural Network for Facial Beauty Prediction.

Performance Comparison

CRNet achieves the state-of-the-art performance on SCUT-FBP and ECCV HotOrNot dataset. For more details, please read our paper.

Performance Evaluation on SCUT-FBP

Methods	PC
Gaussian Regression	0.6482
CNN-based	0.8187
PI-CNN	0.87
Liu et al.	0.6938
CRNet	0.8723

Performance Evaluation on HotOrNot

Methods	PC
Multi-scale Model	0.458
S. Wang et al.	0.437
CRNet	0.482

Test

The most attractive parts learned by CRNet is shown as follows.

Note

A updated version of CRNet with huge improvement can be found from ComboLoss.

Resource

If you find the code or the experimental results useful in your research, please cite our paper:

@inproceedings{xu2018crnet,
  title={CRNet: Classification and Regression Neural Network for Facial Beauty Prediction},
  author={Xu, Lu and Xiang, Jinhai and Yuan, Xiaohui},
  booktitle={Pacific Rim Conference on Multimedia},
  pages={661--671},
  year={2018},
  organization={Springer}
}