APT: Action localization proposals from dense trajectories


The BMVC2015 paper entitled APT: Action localization proposals from dense trajectories by Jan van Gemert, Mihir Jain, Ella Gati and Cees Snoek is now available. This paper is on action localization in video with the aid of spatio-temporal proposals. To alleviate the computational expensive segmentation step of existing proposals, we propose bypassing the segmentations completely by generating proposals directly from the dense trajectories used to represent videos during classification. Our Action localization Proposals from dense Trajectories (APT) use an efficient proposal generation algorithm to handle the high number of trajectories in a video. Our spatio-temporal proposals are faster than current methods and outperform the localization and classification accuracy of current proposals on the UCF Sports, UCF 101, and MSR-II video datasets.

This entry was posted in Science. Bookmark the permalink.

One Response to APT: Action localization proposals from dense trajectories

  1. Cees says:

    The source code for this work is now available at: https://github.com/jvgemert/apt

    We also noted a mistake in our UCF101 ground truth. We made a mistake in the ground truth. We corrected our ground truth and make it available. We updated the scores in the paper, the numbers are different but the conclusions do not change.

Leave a Reply

Your email address will not be published. Required fields are marked *