T2INFT

HKU MSc project about developing a web app which creates NFTs with use of T2I generative models

Check out our Official website.

Table of Contents

• General Info
• Models Used
• NFT website
• Web framework
• Project Status
• Model Installation
• Sample Usage
• Todo
• Acknowledgements

General Information

• This study aims to work out a practical web application that meets potential NFT buyers' requirements in making their customized NFTs, which uses the Generative model to convert text description to semantically consistent NFT style images. Then the images would be converted into NFT products on Ethereum.

Models Used

• HuggingFace Pretrained Stable diffusion model
• Pytorch implementation StyleGAN2-ADA trained on public kaggle NFT dataset
• ImageMixer https://github.com/kirigiricloud/stable-diffusion

NFT website

Ethereum

Web framework

Frontend : Vue 3 + JavaScript

Backend : Nodejs + MySQL

Project Status

Project is: Done

Model Installation

Python version >= 3.7

cd stable-diffusion
pip install -r requirements.txt

Sample Usage

python main-sd.py prompt user_name
python main-mixer.py user_name

prompt: String, text requirment user_name: String, provided user name for registered user

python main-sd.py 'a photo of an astronaut riding a horse on mars' test_user
python main-mixer.py test_user

Todo

• building skeleton of web app
• choose stable diffusion model to use
• preprocess NFT dataset
• train StyleGAN2
• application basic functionalities
• image-mixer: adjust the file to make it work in local
• back-end: wallet
• front-end: profile
• front-end: login/register
• back-end: mint
• Model: create usable Python script - Stable Diffusion
• Model: create usable Python script - image-mixer
• Connect front-end and back-end
• System testing

Acknowledgements

• This project is inspired by HKU MSc Computer Science Department
• Thanks to all team members