Visual Ambigrams

This is the updated version of Ambifusion. Our proposed model can generate ambigram images from two different text conditions. You can see the details of ambifusion(ver.1) in this paper: ArXiv.

Abstract

In the flowing experiments, the pretrained large-scale Text2Img-diffusion models are used in our proposed ambigram generation modules.
Therefore, The model can generate a variety of ambigram images, not only alphabet letter pairs but also diverse image pairs.
In the following demo codes, we use deep-floyed/IF models as generation modules in the reverse process.

Generated Examples.

♦ Alphabet pair ambigrams

♦ "A↕Y" ambigrams with different styles

♦ Image pair ambigrams

Experiment Setup

Set up diffusers environment.

Demo

Gradio

1. Start gradio web app as following.

python demo.py

2. Access 127.0.0.1:11111 with your web browser.

Make ambigrams

1. Set TestConfigs in ambigram_sample.py.
2. Run the sampling code as following.

python ambigram_sample.py

Make ambigrams (higher resolution)

1. Set TestConfigs in ambigram_sample_hr.py.
2. Run the sampling code as following.

python ambigram_sample_hr.py

Customize

You can change base generation model to other models such as StableDiffusion.
But you might change some of the codes in the ambigram_pipeline.