Spring 2018 ULab Computer Vision

Please fork this repository.

Running on your own computer

1. Fork this github repository (make sure git is setup!)
2. git clone [your forked repo url]
3. cd sp18-ulab-computervision
4. git remote add ulab https://github.com/berkeley-ulab/sp18-ulab-computervision.git
5. Setup Conda via this link
6. cd setup and execute bash setup.sh.
7. Run source activate ulab.
8. Run jupyter lab in this repository.

Updating Your Forked Repository

1. git pull ulab master

Navigating Jupyter Notebooks

This link has a quick start on Keyboard navigation shortcuts.

Curriculum

• Numpy
  • Manipulating matrices and vectors
  • Indexing
  • Broadcasting
• Matplotlib
  • Plotting images
  • Line/scatter plots
• Neural Networks/Deep Learning
  • What is a neural network?
  • Building out your own
• PyTorch
  • Creating and manipulating Tensors - very similar to Numpy
• PyTorch Details: Automatic Differentitation
  • Calculating Gradients
• (In Progress) Conv Neural Networks
  • ...
• Gradient Descent
  • Understanding Loss Functions
  • Understanding Loss functions for Style Transfer
  • Understanding Gradient Descent
• Machine Learning (basics)
  • Data - how data is stored and accessed - with particular focus on images
  • Model
  • Training/Testing
  • Train a model on MNIST

Order of notebooks

0. NeuralStyleTransfer (demo)
1. numpy_matplotlib
2. Neural Networks
3. pytorch
4. autograd_tutorial
5. (In progress) Convolutional Neural Networks
6. Gradient Descent