Skip to content

Latest commit

 

History

History
56 lines (35 loc) · 1.67 KB

File metadata and controls

56 lines (35 loc) · 1.67 KB

Facial Beauty Analysis using Distribution Prediction and CNN Ensembles

Overview

Perform Facial Beauty Prediction (FBP) using CNN ensembles originally trained for face verification (VGGFace2). Earth Mover Distance, Categorical Cross Entropy and Mean Squared Error are used to train models to predict the entire facial beauty distribution.

Table of Contents

Quickstart

Run our facial beauty prediction models on your own images in Google Colab/demo environment.
Open All Collab

Installation

We recommend using a conda environment with python 3.9 to run this project.

  1. Install the dependencies
pip install -r requirements.txt
  1. Download the models from https://drive.google.com/drive/folders/1kShZn7FdNIVBePsOylSF69Svemhg5yRo?usp=sharing. Save in a folder models inside the project directory.

  2. (For training only) Download the MEBeauty dataset from https://github.com/fbplab/MEBeauty-database

Usage

Instructions for usage.

Training

Instructions for training.

Results

Results here.

Citation

If you use this code for your research, please cite our paper.

@article{
}

Note: The model weights we provide can only be used for non-commercial research purposes as the dataset they're trained on (MEBeauty) comes with these rules.

For any questions about these models or paper, please contact the authors by sending email to ahmedamanibrahim@gmail.com