Fast neural style transfer twitter bot. Tweet an image, and it responds with a stylized rendering
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples
output_images
.DS_Store
.gitignore
README.md
input_image.png
style_transfer.sh
twitterbot.py

README.md

style-twitter-bot

🌃 Reimagine any image in the style of Van Gogh's The Starry Night! 🌕

This is a twitterbot that performs the following steps in order once initialized:

- listen for tweets targeted at a given twitter account
- pull the image from the tweet 
- perform neural style transfer on the input image
- reply to the tweeter with the output image in a reply tweet (also saves output to local disk)

Examples

Here are examples from my own bot, implemented with this code as-is. To see more examples, you can visit the bot @StarryNightBot.

note Due to recent changes in the Twitter dev API 😞, my bot is no longer live, so tweets to it won't return anything. Sorry!

Baseball Stadium Example Baseball Stadium Example

Stanford University Example Stanford University

Star Wars Example Star Wars Poster

Install

  • For Ubuntu, $ pip install Pillow in place of PIL. No changes to import necessary.
  • For Ubuntu, $ chmod a+x style_transfer.sh to allow the shell script to run

Run

$ python twitterbot.py

Credit

Uses the starry_night.t7 weights file found in the fast-neural-style repo from the paper Perceptual Losses for Real-Time Style Transfer and Super-Resolution