WiSe 2016/17

Deep Learning for Visual Data

Ph. D. Chuan Li; Univ.-Prof. Dr. Michael Wand

Shortname: 08.079.299
Course No.: 08.079.299

Contents

Understanding visual data is core to many real world applications. Over the last few decades we've seen a blossom of techniques in the fields of machine learning and computer vision. For some recognition tasks, we already have machine intelligence that surpasses human performance (such as face identification). For many others, we have also see exciting progresses, such as voice assistants and self-driving cars. In this lecture we will study some landmark works that reshaped the way machines understand visual data.

This lecture will cover the convergent field of computer vision and machine learning, with focus on "deep neural networks", which is a class of techniques that has led to several major breakthroughs in pattern recognition in the past few years.

We will first have a series of lectures to introduce basics of machine vision (feature, classifier and optimization). Then we will look at deep neural networks with its recent development including CNN, RNN and reinforcement learning. We will get to explore these powerful tools for both discriminative tasks (recognition) and generative tasks (synthesis).

We will have both theoretical and practical tutorials.

Requirements / organisational issues

This lecture will be taught in ENGLISH

Prerequisites

  1. Basic knowledge of Image processing: We focus on image understanding, and expect you to know basics of pixelation, linear filtering etc before the lecture starts.
  2. Calculus, Linear Algebra, Basic Probability and Statistics: You should be comfortable taking derivatives and understanding matrix vector operations and notation. You should also know basics of probabilities, Gaussian distributions, mean, standard deviation, etc. Additional knowledge in computer vision and pattern recognition will be helpful too.

Recommended reading list


To get background knowledge for deep learning, we encourage you to have a look at the Stanford CS231n course http://cs231n.stanford.edu/. It is super useful. 

We also encourage you to check what is going on in the exciting research fields of deep learing. The following links are places to look for high quality research works. 
Awesome Deep Vision: https://github.com/kjw0612/awesome-deep-vision
CVPR: http://www.cv-foundation.org/openaccess/CVPR2016.py
ICCV: http://www.cvpapers.com/iccv2015.html
ECCV: http://www.cvpapers.com/eccv2014.html
NIPS: https://papers.nips.cc/
SIGGRAPH: http://kesen.realtimerendering.com/

 

Dates:

Date (Day of the week)TimeLocation
10/27/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
11/03/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
11/10/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
11/17/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
11/24/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
12/01/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
12/08/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
12/15/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
12/22/2016 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
01/12/2017 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
01/19/2017 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
01/26/2017 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
02/02/2017 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik
02/09/2017 (Thursday)12.00 to 14.0704 426
2413 - Neubau Physik/Mathematik

Semester: WiSe 2016/17