/Generative-Models-Seminars-MIPT

Seminars on master's course "Generative Models in Computer Vision" from ABBYY chair at MIPT

Primary LanguageJupyter NotebookMIT LicenseMIT

Generative-Models-Seminars-MIPT

Seminars on master's course "Generative Models in Computer Vision" from ABBYY chair at MIPT (spring 2020)

Seminars

Module 1 (likelihood) Open In Colab

  • Likelihood model in 1D - fitting histogram using SGD
  • Deep Autoregressive model in 2D
  • Detaled study of RealNVP model in 2D

Module 2 (latent variable models) Open In Colab

  • VAE on 2D data
  • investigation of posterior collapse situations
  • VAE on image data

Module 3 (generative adversarial networks) Open In Colab

  • GAN Loss and Non-saturating Loss on 1D data
  • SNGAN on CIFAR10
  • BiGAN on MNIST
  • CycleGAN on Colored MNIST