Fixed - Fast & Furious with no men and no cars

2019 Dec

The trailer of Fast & Furious with no men and no cars :)
The idea was to revise the original trailer by making only use of machine learning libraries and to not edit anything manually. I ended up using the following pipeline:

  1. detectron2: classification of humans / cars / undefined; mask creation of the areas of interest

  2. age-gender-estimation: gender estimation of areas of interest

  3. generative-inpainting: removing all men / cars; patching up these holes by “hallucinating” content from the remaining pixels

There is also related technical in depth tutorial how to get these machine learning libraries running:

A Friendly ML and Docker Tutorial

The experiment was inspired by the Bechdel Test, which is a measure of the representation of women in fiction like movies and novels. “It asks whether a work features at least two women who talk to each other about something other than a man”.
This experiment was an outcome of my creative residency at the Frank-Ratchye STUDIO for Creative Inquiry.

People only

Inverse – men and cars only

Related

Teaching

Cover image for Get ML Stuff Running, Using Docker – A Friendly Tutorial for Artist and Designers

Get ML Stuff Running, Using Docker – A Friendly Tutorial for Artist and Designers

2020–3

Github tutorial/article, outcome of my residency at the Frank-Ratchye STUDIO for Creative Inquiry, Carnegie Mellon University, Pittsburgh, US

Professional-experience

Cover image for Creative Resident at the Frank-Ratchye STUDIO for Creative Inquiry / Visiting Faculty at the School of Design, Carnegie Mellon University

Creative Resident at the Frank-Ratchye STUDIO for Creative Inquiry / Visiting Faculty at the School of Design, Carnegie Mellon University

2019–8

Creative Resident at the Frank-Ratchye STUDIO for Creative Inquiry and Visiting Faculty (for Data Viz) at the School of Design, Carnegie Mellon University, Pittsburgh, US