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

Experimental evaluation

Work package leader: MVL

Model structure is only the first step towards the working system for action recognition. Such complex models usually include a large number of parameters, and the performance of the model in a particular task depends heavily on the values of those parameters. Those parameters can be seen as hidden variables of the compositional model. Therefore, the only way to obtain values of those parameters, is through the process of identification, by running the model on appropriate data and observing its output. The identification of those parameters is usually performed concurrently with learning. The training set of input data is usually split into two parts, with the larger part used for actual learning, and the smaller part used for evaluation of algorithm’s performance. Therefore, model parameters are essentially adjusted through meta­learning.

When shape­based compositional hierarchical model is used for object categorization [Fidler2010a], the lowest levels of the hierarchy can be determined in advance. Parts on the second level are composed from the very basic perceptual units (Gabor filters) of the first level, however they still represent universal building blocks that are largely independent from the particular recognition or categorization task, and similar can be said for the organization of parts on the second level. This means, that those two layers can be trained independently from the actual training data, and this training can be interpreted as part of a model adjustment process. Additionally, extension of compositional hierarchical model into the spatio­temporal domain will undoubtedly increase model’s complexity, and parameter adjustment will be even more challenging.

Therefore, this work package will be composed 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.