Tensorflow and Deep Learning without a PhD - Session 1: CNN

GDG Seattle
Sat, Jan 27, 2018, 10:00 AM (PST)

About this event

Please register for the event on Eventbrite here (https://www.eventbrite.com/e/tensorflow-deep-learning-without-a-phd-session-1-cnn-tickets-42057759892?discount=GDGSeattle). RSVP is closed on this meetup page to avoid confusion.

Come join a TensorFlow and Deep Learning crash course designed for developers and deep learning beginners! In this first session of TensorFlow and Deep Learning without a PhD, Martin Gorner from Google will teach us dense and convolutional neural networks.

In the first talk, you will learn the basic ingredients of dense and convolutional neural networks, how to train and optimize them and how to create a neural network using low-level TensorFlow.

In the second talk, Martin will go over modern advances in convolutional network architectures for classification and detection models. You will learn the architectures of these networks: Inception V3, SqueezeNet and YOLO (for object detection). In contrast to the first talk, you will see how to create a neural network using high-level TensorFlow.

From the 3 hour hands-on codelab in the afternoon, you will learn how to build and train a neural network that recognize handwritten digits, using TensorFlow. if you are attending the codelab, please make sure to complete the installations of Python, TensorFlow and Matplotlib following the instructions here (https://github.com/martin-gorner/tensorflow-mnist-tutorial/blob/master/INSTALL.txt). You will also need to bring your own computer for the codelab.

Event schedule:

9:30 Registration and breakfast
10:00 Talk by Martin Gorner - Tensorflow and Deep Learning without a PhD
11:00 Talk by Martin Gorner - Tensorflow, Deep Learning and Modern Convnets without a PhD
12:00 Lunch
13:00 Codelab - Tensorflow and deep learning without a PhD
16:00 Wrap up

*** Note: You may choose to attend either the morning talks or the afternoon codelab, or both. Space is more limited for the afternoon codelab.