Masterpraktikum KI (D)

Deep Learning for Environment Perception
Dr. M. Lauer

• Eintrag im Vorlesungsverzeichnis (Praktikum)
• Ankündigungen:
  • erstes Treffen: 23.10.2013

• Termine
  • Mittwochs, 12:30 - 14:00, Geb. 082 - SR 00 029

• Sprache
  • Englisch

The lab course on deep learning architectures is intended to get familiar with a new paradigm in machine learning, the so-called deep learning architectures. They are based on neural networks and overcome the problems of classical learning algorithms for neural networks by combining supervised and unsupervised techniques. During recent years amazing results have been achieved with these techniques. Application areas can be found mainly in pattern recognition.

The lab course will start with a short seminar phase in which the participants read and discuss research papers on the topic of deep learning architectures to get a basic idea of this technique. After that, we will apply deep learning techniques to pattern recognition tasks in computer vision and data analysis.

• If you want to roll your own torch distribution you can get it from here: torch
• Update I have updated the install script for installing torch on the PC's in the computer pool to the new torch version! Please use this script now
  • if you are logged in to your account you can simply execute the script like this:

• NOTE before you install the new version make sure you have deleted the old torch_student folder (if you installed to ~/ you want to run: rm -rf ~/torch_student). You also need to update your .bashrc according to the output of the script.

• You can find the lua cheat-sheet here
• The lua tutorial we did in the first week of the course can be found here
• The tutorial for the second week can be accessed here
• A short list of important torch packages to check out can be found here!

• the classification results of the first group phase can be found here
• The corresponding script for outputting your predictions can be found here
• the smiley dataset, training examples and test images ZIP. Note that the data are stored in binary mode, i.e. load them without the 'ascii' option.
• NEW the KITTI data set, cf. http://www.cvlibs.net/datasets/kitti/eval_object.php. A lua file with tools to parse the KITTI label files can be found here ZIP
• NEW a lua file with a function for non-maxima suppression for KITTI objects ZIP

• the full test-dataset, including the labels can now be found here if you want to verify your results

• introduction PDF
• papers for seminar phase PDF
• first working phase PDF
• second working phase PDF
• third working phase PDF

• group 1 (deep Boltzmann machines) PDF
• group 2 (convolutional networks) PDF
• group 3 (deep auto encoders) PDF
• group 4 (comparison of different approaches) PDF