Python script to preprocess images of all Pokémon to finetune ruDALL-E

Overview

ai-generated-pokemon-rudalle

Python script to preprocess images of all Pokémon (the "official artwork" of each Pokémon via PokéAPI) into a format such that it can be used to finetune ruDALL-E using the finetuning example Colab Notebook linked in that repo. This workflow was used to create a model that resulted in AI-Generated Pokemon that went viral (10k+ retweets on Twitter + 30k+ upvotes on Reddit)

My modified Colab Notebook that I used to finetune the model on Pokémon is here: this Notebook's release is purely for demonstration/authentication purposes and no support will be given on how to use it because it is incredibly messy and embarrassing, but there may be a few ideas there that are useful for future generation. Some notes on how the process works are included below, with oppertunity to reproduce/improve it.

The script outputs two things: an images folder with all the preprocessed images plus a data_desc.csv file which contains the image path and Russian caption pairs for finetuning. Some examples of the preprocessed input images are present in the images folder, plus the final data_desc.csv.

The model used is not included in this repo because it's currently too large (~3GB) to distribute (will add the model to Hugging Face at some point).

Preprocessing Script Notes

  • The GraphQL interface to PokéAPI is used as it allows to retrieve the type information plus IDs of all Pokémon in a single request. As a bonus, the returned IDs include the alternate forms of Pokémon (e.g. Mega) which would not otherwise be present just by incrementing IDs.
  • ruDALL-E requires 256x256px, RGB input images. In this case the source input images from PokéAPI are conveiently both square and larger than 256x256 so they downsample nicely. Since the images have transparency (RGBA), they are composited onto a white background.
  • The translation service used is Yandex, which apparently has decent rate limits, plus as a Russian company the translations from English to Russian should theoetically be better.
  • The captions (which are later translated into Russian) are determined by type. For example, a Grass/Poison type will have the caption A Grass-type and Poison-type Pokémon, which is then translated into Russian. In theory, this improves the finetuning process by allowing ruDALL-E to notice trends, plus in theory this can be leveraged at generation-time to control the generation (e.g. prompt with A Grass-type Pokémon and have ruDALL-E generate only Grass-type Pokémon)
  • Due to potential rate limits on translation, translations are cached at runtime by Pokémon type(s) so the API is pinged only once.

Finetuning and Generation Notes

  • The model used above was trained for 12 epochs (4.5 hours on a P100), at a max learning rate of 1e-5. The pct_start param of the OneCycleLR scheudler was set to 0.1 so that learning rate decay happens faster. Despite that, the model converged quickly.

  • The parameters for finetuning ruDALL-E are very difficult to get the expected results. Too little training and the output images will be too incoherent; too much training and the model will overfit and output the source images, and also ignore any text prompts. In the social media posts above, the model is slightly overfit and attempts at using text prompts to control generation failed. But overfitting is not necessairly a bad thing as long as it avoids verbatim output.

Usage

You can install the dependences via:

pip3 install Pillow requests translatepy tqdm

Then run build_image_dataset.py

Getting the images into the ruDALL-E finetuning Colab Notebook is up to the user, but the recommended way to do so is to ZIP the generated images folder (~42 MB!), upload it to Colab (or upload to Google Drive and copy it into the Notebook from there), and unzip the folder in Colab itself via !unzip.

Maintainer/Creator

Max Woolf (@minimaxir)

Max's open-source projects are supported by his Patreon and GitHub Sponsors. If you found this project helpful, any monetary contributions to the Patreon are appreciated and will be put to good creative use.

License

MIT

Owner
Max Woolf
Data Scientist @buzzfeed. Plotter of pretty charts.
Max Woolf
A nonebot2 plugin, send news information in a picture form.

A nonebot2 plugin, send news information in a picture form.

幼稚园园长 7 Nov 18, 2022
A simple, light-weight and highly maintainable online judge system for secondary education

y³OJ a simple, light-weight and highly maintainable online judge system for secondary education 一个简单、轻量化、易于维护的、为中学信息技术学科课业教学设计的 Online Judge 系统。 Onlin

20 Oct 04, 2022
Backtest framework based on DAGs

MultitaskQueue It's a simple framework based on three composed concepts: Task: A task is the smaller unit of execution or simple a node in the DAG, ev

4 Dec 09, 2021
The purpose is to have a fairly simple python assignment that introduces the basic features and tools of python

This repository contains the code for the python introduction lab. The purpose is to have a fairly simple python assignment that introduces the basic

1 Jan 24, 2022
automate some stuff so I can be more noob

dota automate some stuff so I can be more noob This is a simple project, but one that I've wanted forever! I use pyautogui, time, smtplib and datetime

Aaron Allen 17 Oct 18, 2022
Semester long, web application project for CSCI 4370/6370 (Database Management)

Database_Project Prototype ideas for website: Computer Science library (Sells books, products, etc.) Code editor Graph visualizer / creator (can save

Jordan Harman 4 Feb 17, 2022
pyshell is a Linux subprocess module

pyshell A Linux subprocess module, An easier way to interact with the Linux shell pyshell should be cross platform but has only been tested with linux

4 Mar 02, 2022
Larvamatch - Find your larva or punk match.

LarvaMatch Find your larva or punk match. UI TBD API (not started) The API will allow you to specify a punk by token id to find a larva match, and vic

1 Jan 02, 2022
Verification of Monty Hall problem by experimental simulation.

Verification of Monty Hall problem by experimental simulation. |中文|English| In the process of learning causal inference, I learned about the Monty Hal

云端听茗 1 Nov 22, 2022
A Guide for Feature Engineering and Feature Selection, with implementations and examples in Python.

Feature Engineering & Feature Selection A comprehensive guide [pdf] [markdown] for Feature Engineering and Feature Selection, with implementations and

Yimeng.Zhang 968 Dec 29, 2022
Backend/API for the Mumble.dev, an open source social media application.

Welcome to the Mumble Api Repository Getting Started If you are trying to use this project for the first time, you can get up and running by following

Dennis Ivy 189 Dec 27, 2022
A password genarator/manager for passwords uesing a pseudorandom number genarator

pseudorandom-password-genarator a password genarator/manager for passwords uesing a pseudorandom number genarator when you give the program a word eg

1 Nov 18, 2021
Python Cheat Sheet

Introduction Pysheeet was created with intention of collecting python code snippets for reducing coding hours and making life easier and faster. Any c

CHANG-NING TSAI 7.5k Dec 30, 2022
En este repositorio pondré archivos graciositos de python que hago de vez en cuando

🐍 Apuntes de python 🐍 ¿Quién soy? 👽 Saludos,mi nombre es Carlos Lara. Pero mi nickname en internet es Hercules Kan. Soy un programador autodidacta

Carlos E. Lara 3 Nov 16, 2021
A Python 3 client for the beanstalkd work queue

Greenstalk Greenstalk is a small and unopinionated Python client library for communicating with the beanstalkd work queue. The API provided mostly map

Justin Mayhew 67 Dec 08, 2022
Boot.img patcher for Tolino ebook readers to enable ADB and root.

I'm not responsible for any damage to your devices by running this tool. Please note that you may loose warranty when using this, although (This is no

Aaron Dewes 9 Nov 13, 2022
1. 네이버 카페 댓글을 빨리 다는 기능

naver_autoprogram 기능 설명 네이버 카페 댓글을 빨리 다는 기능 네이버 카페 자동 출석 체크 기능 동작 방식 카페 댓글 기능 기본 동작은 주기적인 스케쥴 동작으로 해당 카페 ID 와 특정 API 주소로 대상이 새글을 작성했는지 체크. 해당 대상이 새글 등

1 Dec 22, 2021
Step by step development of a vending coffee machine project, including tkinter, sqlite3, simulation, etc.

Step by step development of a vending coffee machine project, including tkinter, sqlite3, simulation, etc.

Nikolaos Avouris 2 Dec 05, 2021
script buat mengcrack

setan script buat mengcrack cara install $ pkg install upgrade && pkg update $ pkg install python $ pkg install git $ pip install requests $ pip insta

1 Nov 03, 2021
Demodulate and error correct FIS-B and ADS-B signals on 978 MHz.

FIS-B 978 ('fisb-978') is a set of programs that demodulates and error corrects FIS-B (Flight Information System - Broadcast) and ADS-B (Automatic Dep

2 Nov 15, 2022