CoolFace
Apppublic

nmorgan8/ArtisticFaces

sourceHugging Faceupdated 2y agoView on Hugging Face
1likes
App README

ArtisticFaces

Overview

This repo contains code for a machine learning based platform for transfering historical art styles like impressionism and realism into a user's uploaded/captured images in real-time using two different appraaches:

  • —Utilizing layers in VGG - 16 to capture style loss and information for style transference
  • —Training and fine-tuning a Cyclical Generative Adversarial Network (CycleGAN) on ImageNet and Wikiart dataset

Project Structure

  • —src: contains run.py file to launch gradio interface and interact with our application
  • —scripts: contains CycleGAN and VGG-16 related code and dependencies
  • —notebooks: contains .ipnyb notebooks for debugging and demo purposes
  • —training: training script for CycleGAN
  • —data: contains compressed files to image data used to train

Getting Started

To get a local copy up and running, please follow these simple steps.

Prerequisites

Here is what you need to run ArtisiticFaces.

  • —Python3
  • —pip
  • —Miniconda\* (optional to use GPU on Windows machine)

Setup Repository

  1. 1.git clone this repo
git clone https://github.com/nmorgan8/ArtisticFaces.git
  1. 1.cd into directory
cd ArtisticFaces

Setup Python Modules

  1. 1.Install project requirements and set module dependencies
pip install -r requirements.txt
  1. 1.Install tensorflow_examples
git clone https://github.com/tensorflow/examples.git
  1. 1.Set module dependencies
python3 setup.py install

Extract Model Checkpoints

  • —Untar PROD.tar.gz to retrieve trained models
tar -xvzf checkpoints/PROD.tar.gz -C checkpoints/
  • —Unzip PROD.zip to retrieve trained models
unzip checkpoints/PROD.zip
  • —Use softwares (7zip, Archive Utilities, etc.) to extract PROD directory containing the trained models
ArtisticFaces/checkpoints/

Launch Gradio

  1. 1.Navigate to ArtisticFaces home directory
  1. 1.Run src/run.py file to start gradio server
python3 src/run.py

Optional Environment Setup

Optional instuctions to setup GPU if on Windows Machine (tensorflow does not allow MAC GPU use) and create a python virtual environment

Windows GPU Setup

Use the following instructions to setup Windows Machine to run Tensorflow on local GPU

  1. 1.Install and update Nvidia GPU Drivers
  1. 1.Install Miniconda
  1. 1.Create conda environment
conda create --name tf python=3.10
  • —Activate conda evironment
    conda activate tf
  • —Deactivate conda environment
    conda deactivate

Create python virtual environment

Use the following instructions to create a python virutal environment to ensure no dependency issues with python modules.

If using virtual environment, must create env BEFORE installing python modules

  1. 1.Navigate to ArtisticFaces directory
  1. 1.Create Virtual Environment
python3 -m venv <name_of_virtual_env>

Common Errors

  1. 1.ValueError: Trying to load a model of incompatible/unknown type. <file_name>
  • —Delete cached model files to allow it to re-download
rm -r <file_name>
  1. 1.Navigate to ArtisticFaces home directory
  1. 1.Re-launch gradio server

Team Members

  • —Nick Morgan
  • —Claudia Gusti

Technologies

  • —CycleGANs, VGG-Net
  • —Tensorflow, Keras, matplotlib, Gradio

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference