The MechE Alliacne has partnered with MathWorks to bring a series of 3 seminars to the MIT community. 

This first seminar is on the topic: From Pixels to Features to Models: Image Processing, Computer Vision, and Machine- and Deep Learning with Images in MATLAB.  The seminar will be lead by James Wiken who is an application engineer at MathWorks and an MIT Alum.

Transitioning image models from pixel-based to feature-based allows us to extract information from images and video at a high level, to detect, classify, and track objects, co-register images, or understand a real-world scene. Using collections of features, we can train computers to recognize objects, with user-specified or automatically determined features. In this session, you will learn about capabilities for image processing and computer vision with the MATLAB product family. We will also describe approaches to implementing different models for machine learning, including using deep convolutional neural networks (CNNs).

Demos include:

•    Perform automatic image registrations using feature-based approaches

•    Detect, recognize, and track objects in images

•    Ground truth images and video, and train new classifiers

•    Using pre-built networks to create new recognizers in a “transfer-learning-based workflow”

Future seminars in this series include:

• Automating Test Hardware Control and Analysis with MATLAB
• Solving Optimization Problems in MATLAB