Weinland, DanielÖzuysal, MustafaFua, Pascal2010-06-232010-06-232010-06-23201010.1007/978-3-642-15558-1_46https://infoscience.epfl.ch/handle/20.500.14299/50967Most state-of-the-art approaches to action recognition rely on global representations either by concatenating local information in a long descriptor vector or by computing a single location independent histogram. This limits their performance in presence of occlusions and when running on multiple viewpoints. We propose a novel approach to providing robustness to both occlusions and viewpoint changes that yields significant improvements over existing techniques. At its heart is a local partitioning and hierarchical classification of the 3D Histogram of Oriented Gradients (HOG) descriptor to represent sequences of images that have been concatenated into a data volume. We achieve robustness to occlusions and viewpoint changes by combining training data from all viewpoints to train classifiers that estimate action labels independently over sets of HOG blocks. A top level classifier combines these local labels into a global action class decision.text::conference output::conference proceedings::conference paper