• What we'll do
Transfer learning with Tensorflow (or how to avoid training from scratch like the plague)
We'll get started with Tensorflow with a simple computer vision project that can be adapted to many use cases. We will train a model to recognize different varieties of objects. It is meant for people who have Python experience, perhaps already Machine Learning experience (or at least interest in it :) ) but none at all with Deep Learning frameworks. In this session we won't re-write each training block in Tensorflow code yet, but we will go over essential concepts in an image classification task and what is going on inside each block.
• What we assume
- You are used to coding in Python.
- You have a Linux system (OSX is fine; Windows with Anaconda should be ok, but I cannot guarantee at all).
- You already have a clean Python virtual environment ready (both 2 and 3 will work to avoid a flamewar - but I encourage you to use Python 3 ;-) ).
Wifi may not be the best when we're all there, so come prepared and download these couple of things beforehand...
* Data: We'll use the very common and lovely:
You'll recognise this one from many tutorials :)
* Libraries: Tensorflow.
Tensorflow 1.3+ is required. Python 3 is advised.
Install the requirements:
requirements.txt (Save the lines between === in a file named requirements.txt)
pip install -r requirements.txt
Don't worry if you don't have a GPU or a powerful CPU, the trainings will be quite light and can be run even on a Mac.
Speaker: Irina Vidal Migallón
Irina is an Electrical Engineer turned Computer Vision engineer, seasoned in different industries: from optical biopsy systems in France to surgical planning tools or VR/AR apps in the Berlin start-up scene. She's recently joined Siemens Mobility's growing CV & AI team. Even more than waking up Skynet, she's interested in the limits of Natural Intelligence and its decisions over our data.
• Gender policy
We want to make sure that we have at least 50% female attendees, so we ask our non female attendees to come with a female plus one.
• What to bring
Claim the event and start manage its content.I am the organizer