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

L0: Low level motion extraction

A pretty basic procedure is used to extract motion from the video at the lowest layer of the hierarchy. See the two videos that illustrate how it works.

First, a Difference of Gaussians (DoG) is computed for each image from the sequence, and a multi-scale pyramid is built. In the videos below, this is shown in the top-right quadrant of each video. Next, at every scale, the Non-Maxima-Supression (NMS) algorithm is run, pixel-wise on every DoG image, resulting in one surviving pixel per local window, and compared against a fixed threshold, to discard local maxima appearing on uniform surfaces. This is shown in lower-left part of each video - size of the pixels corresponds to the scale the features were detected on. The detections are associated across the neigbouring frames, to form motion vectors (shown in lower-right part of the video, as red lines).

Note that the videos are in high definition and should be played in full screen, to see the necessary details. First video depicts a synthetic scene - a textured rotating wheel, made of hollow squares.

The second video shows features, extracted from the collage of videos from our Robowood dataset.

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.