Login Form


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...


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...


Motion hierarchy structure

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


Server crash

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


L1: motion features

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


How good is the extracted motion?

Our motion extraction mechanism is pretty basic. So, how does it visually compare to state-of-the-art optical flow algorithm?

As a baseline, the  lower left part of each video shows the dense TV-L1 optical flow, published as:

Antonin Chambolle, Thomas Pock: A First-Order Primal-Dual Algorithm for Convex Problems with Applications to Imaging. Journal of Mathematical Imaging and Vision 40(1): 120-145 (2011). The code to calculate the flow was obtained from TU Graz.

The lower right part of each video shows the visualization of the "dense" flow, reconstructed from the output of the L0 layer of our hierarchy. The "dense" representation was calculated as a sum of Gaussians responses, where each Gaussian corresponds to one motion vector, the color is determined by the orientation, the intensity by the magnitude, and the variance by the scale, on which the motion vector was obtained.

Note: These results are obtained for visual comparison only, they are not used for the motion analysis. However, it can be seen that the quality of the extracted motion in the lowest (L0) layer of our hierarchy is quite reasonable, given the very basic nature of the used features.


First video shows the flow, obtaned from one clip from the Holywood2 dataset:

And the second shows the flow, obtained from the collage of our Robowood clips.

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.