TFGENZOO 1.2.5

helper of building generative model with Tensorflow 2.x

TFGENZOO (Generative Model x Tensorflow 2.x)

What’s this repository?

This is a repository for some researcher to build some Generative models using Tensorflow 2.x.

I NEED YOUR HELP(please let me know about formula, implementation and anything you worried)

Zen of this repository

We don't want to need flexible architectures.
We need strict definitions for shapes, parameters, and formulas.
We should Implement correct codes with well-documented(tested).

How to use?

By Install

• pipenv

  pipenv install TFGENZOO==1.2.4.post7
  

• pip

  pip install TFGENZOO==1.2.4.post7
  

Source build for development

1. clone this repository (If you want to do it, I will push this repository to PYPI)
2. build this repository docker-compose build
3. run the environment sh run_script.sh
4. connect it via VSCode or Emacs or vi or anything.

Examples

• TFGENZOO_EXAMPLE

• Simple Tutorials

  • What is the invertible layer

    The tutorial about Flow-based Model

  • conditional flow-based model

    How to add conditional input into Flow-based Model for the image generation.

Documents

https://mokkemeguru.github.io/TFGENZOO/

Roadmap

• Flow-based Model Architecture (RealNVP, Glow)
• i-ResNet Model Architecture (i-ResNet, i-RevNet)
• GANs Model Architecture (GANs)

Remarkable Backlog

Whole backlog is here

News [2020/6/16]

New training results Oxford-flower102 with only 8 hours! (Quadro P6000 x 1)

data	NLL(test)	epoch	pretrained
Oxford-flower102	4.590211391448975	1024	---

see more detail, you can see my internship’s report (Japanese only, if you need translated version, please contact me.)

News [2020/7/11]

Add some tutorial into ./tutorial

News [2021/3/30]

I wrote the master's paper about japanese text style transfer. "AutoEncoder に基づく半教師あり和文スタ イル変換" https://drive.google.com/file/d/1KtkLZi6PUvL7msAqbg_KRdEC0pmmpbhf/view?usp=sharing

Contact

MokkeMeguru (@MokkeMeguru): DM or Mention Please (in Any language).