From the recent ENS/INRIA Summer school on Visual Recognition and Machine Learning that took place in Paris, 25-29 July 2011. I note that Jean Ponce makes the point that sparse coding is not compressed sensing, a little bit like what Muthu was saying a while back. Obviously, becoming mainstream has its downs, but let me wonder something aloud. The whole field of computer vision is based on the underlying assumption that the transfer function of most camera equipment between a scene and its 2-D projection on the focal plane array is a simple one. What if it were a more complex one ? Would most of the techniques used in computer vision and image understanding also work or break down ? To me this is an essential question.
Here are the slides of the meeting, the materials for the practical sessions is here. Enjoy the learning experience.
Lecture slides
- Supervised learning, SVMs, kernel methods, Francis Bach
- Instance-level recognition (part1, part2), Cordelia Schmid and Josef Sivic
- Large-scale visual search (part1, part2), Cordelia Schmid and Josef Sivic
- Bag-of-Features models for category-level classification, Cordelia Schmid
- Sparse coding and dictionary learning for image analysis, Jean Ponce
- Category-level localization, Andrew Zisserman
- Learning with structured inputs and outputs, Christoph Lampert
- Attributes for object recognition, Christoph Lampert
- Recent progress on visual recognition, Jitendra Malik
- The future of visual recognition, Jitendra Malik
- Large scale learning, Léon Bottou
- Large-scale visual recognition I, Alyosha Efros
- Large-scale visual recognition II, Florent Perronnin
- Human actions, Ivan Laptev
- Using geometric information in recognition and scene analysis, Martial Hebert
Liked this entry ? subscribe to the Nuit Blanche feed, there's more where that came from
No comments:
Post a Comment