These are all costumes generated by a neural network trained by Jacobs School alumna Janelle Shane. Shane, who earned a Ph.D. in electrical engineering at UC San Diego in the lab of Professor Shaya Fainman, works with lasers by day. But her hobby is working with neural networks to create funny data sets.
For this project, she crowdsourced 4500 costume ideas from her blog readers and fed them to a neural network.
The network did not disappoint, generating costume suggestions such as vampire Big Bird, celery blue Frankenstein and strawberry shark.
Soon, Shane's readers were getting into the game and drawing the costumes the neural network suggested.
First up, strawberry shark:
And then: Bearley Quinn (courtesy of Twitter user @vonbees):
But Shane's readers weren't done. Soon they started making some of the costume suggestions a reality.
Twitter user Liz Walsh dressed up as the Dragon of Liberty:
Twitter user @HerbLovesTech and his wife dressed up as Professor Panda and Shark Princess:
And Shane? She took her inspiration from an entry in the costume data base. She will be Ruth Vader Ginsburg (that's a mash up of Supreme Court Justice Ruth Bader Ginsburg and Star Wars villain Darth Vader):
For more neural-network generated Halloween costumes, read Shane's blog post here. And read this news story by writer Rae Paoletta here and this Popular Mechanics story by writer Sophie Weiner.
No comments:
Post a Comment