• Join remotely here: https://meet.google.com/qij-pjqj-acr
• What we'll do:
This is a hands-on workshop on machine vision. During this workshop, attendees will learn: 1) What is a convolutional neural network by building one, 2) Transfer learning, 3) Object prediction and building object prediction pipelines. We will build a neural network from scratch and train it to recognize images. Once we create a network, we will improve this model using a strategy called transfer learning. Once we understand the concepts of transfer learning, we will train a deeper network called Inception version 3 to create a state of the art image classifier. The projects will use OpenCV, Tensorflow and Keras; three very popular machine vision and deep-learning tools. We will also cover the basics of deploying scalable python applications in the cloud.
The topic for this week is explainable deep learning in the form of visualizing the predictions. To achieve this goal, we will be using a technique called class activation maps.
The course is hosted using either our own cloud platform: Jomiraki, a cloud connected AI developer environment or Google CoLab, a GPU powered Jupyter compatible deep-learning instance. Either of these environments will be set-up ahead of time, with zero end-user dependencies. This will ensure that each participant will spent more time testing and running the code, instead of trying to figure out the set-up process itself.
• What to bring:
This a bring your own device (BYOD) event. For optimal experience, Moad machine vision team recommends Chrome >=72, to access the course contents.
Please set-up a Kaggle and GitHub account ahead of time. Both accounts are needed to get the full benefit of the code examples.
• Important to know:
This is an introductory workshop on machine vision. This course is part of the FutureReady boot-camp by Moad Computer. If you are interested in participating in the boot-camp, please fill-out this form: https://goo.gl/forms/TzClAtTqOLwHcudv1
• Additional materials: