Online
Dec 7, 9 &13, 2022
1:00 - 4:00 pm EST
Instructors: Zachary Frohock
Helpers: Jim Peterson, Rohit Tripathy, Hong Wang
Some adblockers block the registration window. If you do not see the registration box below, please check your adblocker settings.
Machine learning has been used by computational biologists for decades, but only more recently have developments in the field allowed for the timely and accurate processing of image data. By leveraging and augmenting the skills of domain experts, classification of images and sub-regions is not only possible, but fast and reliable. This course serves as an introduction to the application of machine learning to image segmentation, making use of two powerful, free-to-use components: 1. Tensorflow 2.x – an open source machine learning platform used in research, production, and everything in between, 2. Google Colaboratory – a browser based utility to run Python code with minimal setup required. We’ll use these components in a selection of hands-on demos that could easily be adapted to work with participants’ own data. First, we’ll walk through some machine learning image processing history. Next, we’ll introduce and implement a convolutional neural network (CNN). Finally, we’ll build upon CNN’s to discuss a more recent development in model architecture, the U-Net. Plan to attend all three sessions in the course.
Who: The course is aimed at students and other researchers. You can participate in person (at Bar Harbor and Farmington) or remote participation will be available. Familiarity with Python to some degree is necessary. Experience with Jupyter Notebooks, numpy, and pandas will be useful.
Where: This training will take place both online and in-person at the Bar Harbor and Farmington campuses. If you will attend online, the instructors will provide you with the information you will need to connect to this meeting.
When: Dec 7, 9 &13, 2022. Add to your Google Calendar.
Requirements: Participants must have access to a computer with a Mac, Linux, or Windows operating system (not a tablet, Chromebook, etc.). They should have a few specific software packages installed (listed below).
Accessibility: We are dedicated to providing a positive and accessible learning environment for all. Please notify the instructors in advance of the workshop if you require any accommodations or if there is anything we can do to make this workshop more accessible to you.
Contact: Please email zachary.frohock@jax.org or yi.li@jax.org for more information.
Roles: To learn more about the roles at the workshop (who will be doing what), refer to our Workshop FAQ.
| 1:00 | Intro to the Workshop - Sessions and Covered Topics |
| 1:20 | Brief History of Early Machine Learning Development |
| 2:30 | Break |
| 2:45 | Short Multilayer Perceptron Example - Classifying Handwritten Digits |
| 3:15 | Introduction to Convolutional Neural Networks (CNN) |
| 3:30 | CNN example - Image Classification |
| 4:00 | END |
| 1:00 | Review of CNNs for Whole Image Classification |
| 1:30 | Introduction to the Keras functional API |
| 1:30 | From Whole Images to Pixel-wise Classifications |
| 2:30 | Break |
| 2:45 | Introduction to U-Net Architecture |
| 3:00 | U-Net Example |
| 3:15 | Demo of a Production U-Net model |
| 4:00 | END |
| 1:00 | Review of Demo U-Net Model |
| 1:30 | Model Hyperparameters and Optimization |
| 2:30 | Break |
| 2:45 | Short Selection of Guided Exercises |
| 4:00 | END |