DLCV - Deep Learning for Computer Vision

School

ETSETB Telecom BCN

Program and level

Master in Telecommunications Engineering (MET)

Acronym

DLCV

Contents

In partnership with Insight Centre for Data Analytics - Dublin City University (DCU) and Barcelona Supercomputing Center (BSC).

Dates: 4-8 July, 2016

Location: Campus Nord of the Universitat Politècnica de Catalunya, Barcelona.

Schedule: 3-7pm

Course instructors: Xavier Giró-i-Nieto (UPC), Elisa Sayrol (UPC), Amaia Salvador (UPC), Jordi Torres (BSC), Kevin McGuinness (DCU) and Eva Mohedano (DCU).

The aim of this course is to train students in methods of deep learning for computer vision. Convolutional neural networks (convnets) will be presented and analyzed in detail to understand the potential of these state of the art tools in visual pattern recognition. Engineering tips and scalability issues will be addressed to solve tasks such as image classification, object detection or automatic textual captioning. Hands-on sessions will provide development skills so that attendees can solve a selected task in an open scientific benchmark and, if successful, submit their results. 

 

Content: 

1. Convolutional Neural Networks

Architecture: Forward and recurrent networks.

−Backpropagation

−Layer Visualization.

−Memory and computational requirements.

−Best practices.

−Fine-tunning

 

2. Applications

Image retrieval and classification

−Face and object detection/recognition.

−Semantic segmentation

−Saliency prediction

−Image captioning

−Multimodal fusion 

 

Laboratory practical exercises:

•Description: Training of a convnet for character recognition. (1 hour)

•Description: Visualization and ablation of convnet layers. (1 hour)

•Description: Fine-tunning a convnet for transfer learning. (1 hour)

•Description: Local image analysis. (1 hour)

 

Bibliography:

- Yoshua Bengio, Ian Goodfellow, Aaron Courville, "Deep Learning", MIT Press, In preparation (http://www.deeplearningbook.org/)

- Jordi Torres, "Hello World en TensorFlow", lulu.com (2016)

 

Videos:

- Christof Angermueller and Alex Kendall, "Convolutional Neural Networks". Cambridge University Machine Learning (May 2015).

- Matt Zeiler, "Understanding and Visualizing Neural Networks". Hakka Labs (February 2015).

 

Related Courses:

- Fei-Fei Li, Andrej Karpathy, "CS231n: Convolutional Neural Networks for Visual Recognition". Stanford University 2016. [Videos] 

- Hugo Larochelle, "Neural Networks". Université de Sheerbroke

- Joan Bruna, "Stats212b: Topics on Deep Learning". Berkeley University (Spring 2016).

- Yann LeCun, "Deep Learning: Nine Lectures at Collège de France"Collège de France (Spring 2016). [Facebook page]

- Graham Taylor, Marc'Aurelio Ranzato, Honglak Lee, "Tutorial on Deep Learning for Vision". Intlernational Conference on Computer Vision (CVPR) 2014. 

- German Ros, Joost van de Weijer, Marc Masana, Yaxing Wang, "Hands-on Deep Learning". Computer Vision Center (CVC) 2015.

- Vincent Vanhoucke, Arpan Chakraborty, "Deep Learning". Udacity 2016.

- Jordi Torres, "First contact with TensorFlow". Barcelona Supercomputing Center, 2016.

 

Websites:

- Deep Learning Glossary by WildML