Format

The tutorials consist of lecture segments, demos, and hands-on exercises. We strongly encourage you to bring a laptop with all the required packages installed in order to participate fully.


Data

Some of the tutorials will be using a data-set of cell images kindly provided by Maryana Alegro (UCSF). The data is available to download here.

The Azure tutorial will use an image data-set described in this webpage. Please use the link on the page to download the data.


Software

Python for Science

  • Python

    If you are new to Python, please install the Anaconda distribution for Python version 3 (available on OSX, Linux and Windows). Or, feel free to use your favorite distribution, but please ensure the requirements below are met:

    • numpy >= 1.11
    • scipy >= 0.18
    • matplotlib >= 2.0
    • skimage >= 0.12
    • sklearn >= 0.18
    • dask >= 0.14
    • keras >= 1.2
    • tensorflow >= 1.0
    • notebook >= 4.0

    In Anaconda, install these packages with:

    conda install numpy scipy matplotlib jupyter scikit-image scikit-learn dask
    pip install conda tensorflow  # or tensorflow-gpu if you have an Nvidia graphics card
    

    In any other Python distribution, use pip:

    pip install numpy scipy matplotlib jupyter scikit-image scikit-learn dask keras tensorflow  # or tensorflow-gpu if you have an Nvidia graphics card
    

    In the next section below, we provide a test script to confirm the version numbers on your system.

Test your setup

Please run this test script to verify your package versions.

E.g., on my computer, I see:

$ python check_setup.py
[✓] numpy         1.12.1
[✓] scipy         0.19.0
[✓] matplotlib    2.0.0
[✓] notebook      4.3.1
[✓] scikit-image  0.13dev
[✓] scikit-learn  0.18.1
[✓] dask          0.14.0
[✓] tensorflow    1.0.0
Using Theano backend.
[✓] keras         2.0.2

To update a package, use:

conda upgrade package_name          # for conda
pip install --upgrade package_name  # for pip

If you do not have a working setup, please contact the instructors.

ImageJ

Fiji

Installation instructions are available here

For this demo we will use samples that will be fetched from the imagej.net site through the application, so no separate downloads are required.

ImageJ Notebooks

We encourage downloading Beaker and opening the ImageJ tutorials notebook prior to the tutorial. This will automatically fetch the necessary dependencies.

Recently the ImageJ community has begun using Beaker Notebook for interactive coding. Beaker is like Jupyter notebook, but is built upon the JVM. It has some unique features that are a bit more challenging to incorporate into Jupyter notebook; in particular, Beaker is a polyglot notebook where switching between languages within a single notebook while maintaining access to the same scope/variables.

Beaker installation instructions are available here (only the “base install” is required).


Lecture materials

Numpy, Scipy, Matplotlib, Scikit-image

There are two ways of downloading the lecture materials:

a) Get the ZIP file from GitHub

b) Clone the repository at https://github.com/imagexd/2017-tutorials

ImageJ

We will work through the ImageJ beaker tutorial from the imagej-tutorials github repository, which can either be checked out from the repository: https://github.com/imagej/tutorials

Or downloaded directly with this link

Image Classification using Azure

In this tutorial, we will build an SVM based image classifier using Azure Machine Learning. You will learn how to use Azure storage, get familiar with the Azure portal, and with Azure Machine Learning. Participants will be given $500 worth of Azure credits that will be valid for one month

Please download the latest version of the material on the morning of the tutorials to be sure you have all updates.