Login Form

Editors

Activity recognition results on UCF Sports and Holywood2

Table above shows the results, obtained on UCF Sports dataset (http://crcv.ucf.edu/data/UCF_Sports_Action.php). We report recognition rate with respect to the number...


Read More...

Computational efficiency and parallel implementation

The developed algorithms are computationally effective and the compositional processing pipeline is well-suited for implementation on massively parallel architectures. Many...


Read More...

Motion hierarchy structure

Our model is comprised of three processing stages, as shown in the Figure. The task of the lowest stage (layers...


Read More...

Server crash

After experiencing a total server failure, we are back online. We apologize for the inconvenience - we are still in...


Read More...

L1: motion features

Layer L1 provides an input to the compositional hierarchy. Motion, obtained in L0 is encoded using a small dictionary.


Read More...
01234

Work programme

The result of this project will be the methods for efficient representation of visual motion, which can be used for perception and classification of different motion patterns in artificial systems. We will base our work on the paradigm of hierarchical compositionality. There are several studies (e.g., [Bienenstock1994, Fiser2002]) supporting the theory that it is exactly the compositionality that allows living beings to efficiently perceive and analyze various perceptual categories. Previous research in computer­vision­based shape and motion recognition [Jhuang2007, Fidler2010a] also supports this theory from the computational standpoint. It has been shown in [Fidler2010a] that compositional representation of shapes allows compressed representation of a large number of shape categories, fast inference on input data and semi­supervised learning from large amounts of data.

The development of such a methodology will, however, require synergy from various experts. Having this in mind, we have composed the project consortium from four groups, and the contribution of each is crucial for achieving the project’s goal. The VICOS group has extensive experience with hierarchical compositional representations for shape categorization [Fidler2006a, Fidler2007, Fidler2008, Fidler2009a, Fidler2010a, Fidler2009b] and will provide central insights into the methods for building hierarchies of motion. The MVL group has a long history of work in motion analysis [Perš2002, Perš2003, Kristan2009a, Kristan2009b, Perše2009, Kristan2010b, Perše2010, Perš2010]. They will contribute their expertise in motion tracking and in the methodology for motion representation and recognition. From the implementation standpoint, we expect that parallel implementations of the developed algorithms will be necessary both to achieve top performance and to enable sufficient pace of experimental work. Parallel architectures and algorithms are one of the key competences of the DCS group [Trobec2000, Šterk2005, Depolli2008, Trobec2009a, Trobec2009b], and we will use their expertise in this field to arrive at efficient implementations of the developed methods. As an important part of development of the motion representation methodology, realistic motion datasets, recorded in controlled manner, will be required. The DABR group has a longstanding history in humanoid cognitive robotics [Ude2003, Krueger2010, Ude2010, Ude2008], and will contribute both their expertise and the technology for this task. They will use their humanoid robots to generate the appropriate experimental data.

By achieving our goals we expect significant advancements in the state­of­the­art in motion analysis and activity recognition by our hierarchical compositionality models. With respect to the issues that we have pointed out above (and detailed in the “Scientific background”) and the expertise of each project partner, we have structured the work packages and assigned the leaders of each work package. The workpackages will be structured as follows.

This website uses cookies to manage authentication, navigation, and other functions. By using our website, you agree that we can place these types of cookies on your device.

View e-Privacy Directive Documents

You have declined cookies. This decision can be reversed.

You have allowed cookies to be placed on your computer. This decision can be reversed.