arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1405.7545cs.CV
收藏

大型动作分类数据集上视觉词汇生成的特征采样和划分

Feature sampling and partitioning for visual vocabulary generation on large action classification datasets

Oxford Brookes University(牛津布鲁克斯大学);University of Oxford(牛津大学)

Michael Sapienza, Fabio Cuzzolin, Philip H. S. Torr

摘要:

动作识别的最新趋势是更大的数据集,越来越多的动作类和更大的视觉词汇表。目前,在具有挑战性的视频数据中,最先进的人类动作分类是基于视觉词袋管道,其中时空特征被全局聚合以形成直方图。选择的策略来采样特征和构建视觉词汇对性能至关重要,实际上往往主导性能。在这项工作中,我们提供了一个关键的评估各种方法来建立一个词汇,并表明良好的词汇确实有显着的影响。通过策略性地对特征进行子采样和分区,我们能够使用相对较小的视觉词汇在5个主要动作识别数据集上实现最先进的结果。

英文摘要:

The recent trend in action recognition is towards larger datasets, an increasing number of action classes and larger visual vocabularies. State-of-the-art human action classification in challenging video data is currently based on a bag-of-visual-words pipeline in which space-time features are aggregated globally to form a histogram. The strategies chosen to sample features and construct a visual vocabulary are critical to performance, in fact often dominating performance. In this work we provide a critical evaluation of various approaches to building a vocabulary and show that good practises do have a significant impact. By subsampling and partitioning features strategically, we are able to achieve state-of-the-art results on 5 major action recognition datasets using relatively small visual vocabularies.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑