Home

Welcome to the artistic-styles wiki!

Artistic Styles

About

Add more about the project here

Links

• A Neural Algorithm of Artistic Style
• VGG - Visual Geometry Group
  • Very Deep Convolutional Networks for Large-Scale Image Recognition Paper
  • [Results] http://www.robots.ox.ac.uk/~vgg/research/very_deep/
  • Thank You Toronto
• Paper 2
• Tutorial // Tutorial on how VGG was coded
• CNN Convolutional Neural Networks as described on deeplearning.net website; includes code samples for Theano
• [jcjohnson-neural-style] (https://github.com/jcjohnson/neural-style)
• Lecture on Con-nets// lecture explaining con Nets and their history most useful for us is the section starting at 52m which talks about using a pre-trained con-net eg VGG and appending a new task to the end. Exactly what was done for the paper we are investigating.

Technology

• Theano
• Caffe
• keras

Tools

• tmux tutorial
  • intro : look up the following terms
    • session, windows and panes
    • tmux shortcut
  • basic

  tmux new-session -s session1            # create a new session
  ctrl+b, d                               # detach
  tmux attach -t session1                 # attach to a new session
  ctrl+b, c                               # create a new window
  ctrl+b, [0-9]                           # go to a window
  

  • intermediate
    • go through the reference
    • instead of tmux, use tmux -S /path/to/tmux-shared to share tmux screens (so long as the file /path/to/tmux-shared is accessible to users)

Questions

• Which layers could we extract the information?
  • conv1_1, conv2_1, conv3_1, conv4_1, conv5_1
• Reconstruction of images
  • Model
  • Weights
  • How to train