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

Compositional hierarchy for motion analysis

Work package leader: VICOS

The basic concept of hierarchical compositional structure stems from human visual system. Nevertheless, since the visual system itself is too complex and not sufficiently understood to be directly modeled in software, the computational entities of compositional hierarchy for motion analysis need to be derived anew. We build on the experience from shape categorization, where such models are already available.This task is by no means trivial, as introducing the concept of time into already complex 2D­based structure is more complex than just adding another dimension to the compositional models for object categorization. While Gabor filters seemed to be a natural choice for the shape­based compositional model for object categorization, we treat the selection of lowest level features for action recognition as an open research question. Likewise, we cannot predict what the complexity of motion recognition hierarchy in comparison to object recognition would be, and consequently, it is unclear how many levels in the hierarchy will be actually needed. Thus, these issues will be addressed during the project. Also, while it is clear to which influences the developed algorithms should be invariant to, it is not clear, which level should provide invariance to a particular influence. We will need to deal with the problem of spatio­temporal scale as well, and preferably do it in an efficient way. Finally, given much larger complexity of motion recognition in comparison to object recognition, it is unclear which learning methods and which stochastic models will be appropriate to model the relations among different parts.

The work package WP1 will be comprised of the following tasks:

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.