Building Advanced Image Classifiers with CNNs using PyTorch (Online)

CNNs (Convolutional Neural Networks) have revolutionized the way we build visual processing systems and are a major component of the deep learning toolkit available to machine learning practitioners. CNNs are commonly used in visual processing but also have applications in other fields like Natural Language Processing. In this talk, Jacob will introduce CNNs and give a detailed overview of how they work and some of the motivations behind why there were invented. We will then get our hands dirty and build a CNN model from scratch using PyTorch and train it to recognize different objects from a common visual dataset. We will then use a transfer learning technique to improve the accuracy of our model.

A large part of this talk will be a hands-on lab so do bring your GPU-enabled laptop pre-loaded with Anaconda, PyTorch and Jupyter Notebook. If you don’t have a GPU enabled laptop, Jacob will be sending out a Google Colab option which will allow you to build and train the model in Google’s cloud (for free) using your web browser. Instructions for both will be sent out before the talk. The lab will require you to have some knowledge of Python programming. If you don’t have Python experience, you should still be able to benefit from following along the coding examples.

Bring a laptop to maximize participation.

About the speaker:
Jacob Jensen is a deep learning specialist and currently a mentor at Udacity’s Deep Learning Nanodegree Program working with students to apply state-of-the-art deep learning technology to solve real world business problems. Before, he was the Sr. Director of Product Management and Marketing at Cisco Systems. Jacob has a M.S. in Electrical Engineering from Stanford University and has held multiple leadership positions in engineering, product management and marketing over the past 20 years spanning both large companies and start-ups.

A link will be sent out prior to the event to those who have RSVPd.