Skip to content

EXJUSTICE/StyleGAN-RNN-Tinder_profile_generation

Repository files navigation

Generating Swipeable Tinder Profiles using AI: Adversarial & Recurrent Neural Networks in Realistic Content Generation

An implementation of StyleGAN and RNN architectures to generate profile biographical and image information, with the aim of generating believable female profiles as a proof of concept. All notebooks were ran in Google's Colaboratory environment with GPU acceleration enabled.

  • StyleGAN pre-trained on FFHQ 1024x1024 dataset, cloned from NVIDIA's StyleGAN repository
  • RNN trained on Kaggle Wine Reviews Dataset. Based upon the TrumpChar RNN project, with input from Brownlee's excellent RNN tutorials.
  • RNN trained on the now-deleted Kaggle Tinder Female Profiles dataset. Legacy attempt kept for reference purposes, without access to dataset.

Contents

  • RNN training notebook
  • RNN text generation notebook
  • StyleGAN image generation notebook

Full tutorial available here: https://medium.com/gradientcrescent/generating-swipeable-tinder-profiles-using-ai-adversarial-recurrent-neural-networks-in-dd68bd98c2f3

Releases

No releases published

Packages

No packages published