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

Task 2.1: Experimental database

An important challenge in this task is the choice of the appropriate database, which should contain representative samples of human motion and human activities. We plan to acquire and properly annotate three separate experimental data sets. First data set will contain synthetic motion, performed by humanoid robot. It is important, that the research is started on the well­controlled data set, and the synthetic motion will allow us to increase the complexity of the task in a controlled manner. This data set will be acquired in several stages, starting with basic human gestures and basic activities. Such data has low complexity, and will be used in early stages of development, when models with smaller number of layers will be tested. We will gradually increase the complexity of activities, taking the inspiration from human motion, especially from the well documented and researched human activities, for example, sports. It should be noted, that although the motion will be synthetic, the visual data will be real, recorded using off­the­shelf imaging equipment, to increase the relevance of the data for the actual activity recognition problems.

The second data set will contain actual human motion, and will be structured similarly to the first data set. It will be used to evaluate performance of the model at various stages of development, after we will make sure that the model performs well on the synthetic motion data set. The complexity of motion in that data set will be also gradually increased, but due to practical problems we do not expect to have such fine control over the complexity of motion as in the synthetic data set. Due to practical problems with acquisition of motion of real human subjects,we don’t expect more than two or three levels of complexity. If possible, existing human motion databases will be used as part of the second data set.

The third data set will contain both motion and shape cues and will be used in evaluation of combined shape and motion model. to show the advantage of the combined model. We expect that this data set will be the smallest, however, it will require significant amount of planning, to be relevant for real world problems.

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.