Realtime Video Classification Using Dense HOF/HOG

J. R. R. Uijlings; I. C. Duta; N. Rostamzadeh; N. Sebe

doi:10.1145/2578726.2578744

Realtime Video Classification Using Dense HOF/HOG

J. R. R. Uijlings, I. C. Duta, N. Rostamzadeh, N. Sebe

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

The current state-of-the-art in Video Classification is based on Bag-of-Words using local visual descriptors. Most commonly these are Histogram of Oriented Gradient (HOG) and Histogram of Optical Flow (HOF) descriptors. While such system is very powerful for classification, it is also computationally expensive. This paper addresses the problem of computational efficiency. Specifically: (1) We propose several speed-ups for densely sampled HOG and HOF descriptors and release Matlab code. (2) We investigate the trade-off between accuracy and computational efficiency of descriptors in terms of frame sampling rate and type of Optical Flow method. (3) We investigate the trade-off between accuracy and computational efficiency for the video representation, using either a k-means or hierarchical k-means based visual vocabulary, a Random Forest based vocabulary or the Fisher kernel.

Original language	English
Title of host publication	Proceedings of International Conference on Multimedia Retrieval
Place of Publication	New York, NY, USA
Publisher	ACM
Number of pages	8
ISBN (Print)	978-1-4503-2782-4
DOIs	https://doi.org/10.1145/2578726.2578744
Publication status	Published - 2014

Keywords / Materials (for Non-textual outputs)

Computational Efficiency, HOF, HOG, Video Classification

Access to Document

10.1145/2578726.2578744

realtimeVideoAuthorVersionAccepted author manuscript, 402 KB

http://doi.acm.org/10.1145/2578726.2578744

Cite this

@inproceedings{3920184830d542bc944a31f1965b8936,

title = "Realtime Video Classification Using Dense HOF/HOG",

abstract = "The current state-of-the-art in Video Classification is based on Bag-of-Words using local visual descriptors. Most commonly these are Histogram of Oriented Gradient (HOG) and Histogram of Optical Flow (HOF) descriptors. While such system is very powerful for classification, it is also computationally expensive. This paper addresses the problem of computational efficiency. Specifically: (1) We propose several speed-ups for densely sampled HOG and HOF descriptors and release Matlab code. (2) We investigate the trade-off between accuracy and computational efficiency of descriptors in terms of frame sampling rate and type of Optical Flow method. (3) We investigate the trade-off between accuracy and computational efficiency for the video representation, using either a k-means or hierarchical k-means based visual vocabulary, a Random Forest based vocabulary or the Fisher kernel.",

keywords = "Computational Efficiency, HOF, HOG, Video Classification",

author = "Uijlings, {J. R. R.} and Duta, {I. C.} and N. Rostamzadeh and N. Sebe",

year = "2014",

doi = "10.1145/2578726.2578744",

language = "English",

isbn = "978-1-4503-2782-4",

booktitle = "Proceedings of International Conference on Multimedia Retrieval",

publisher = "ACM",

}

TY - GEN

T1 - Realtime Video Classification Using Dense HOF/HOG

AU - Uijlings, J. R. R.

AU - Duta, I. C.

AU - Rostamzadeh, N.

AU - Sebe, N.

PY - 2014

Y1 - 2014

N2 - The current state-of-the-art in Video Classification is based on Bag-of-Words using local visual descriptors. Most commonly these are Histogram of Oriented Gradient (HOG) and Histogram of Optical Flow (HOF) descriptors. While such system is very powerful for classification, it is also computationally expensive. This paper addresses the problem of computational efficiency. Specifically: (1) We propose several speed-ups for densely sampled HOG and HOF descriptors and release Matlab code. (2) We investigate the trade-off between accuracy and computational efficiency of descriptors in terms of frame sampling rate and type of Optical Flow method. (3) We investigate the trade-off between accuracy and computational efficiency for the video representation, using either a k-means or hierarchical k-means based visual vocabulary, a Random Forest based vocabulary or the Fisher kernel.

AB - The current state-of-the-art in Video Classification is based on Bag-of-Words using local visual descriptors. Most commonly these are Histogram of Oriented Gradient (HOG) and Histogram of Optical Flow (HOF) descriptors. While such system is very powerful for classification, it is also computationally expensive. This paper addresses the problem of computational efficiency. Specifically: (1) We propose several speed-ups for densely sampled HOG and HOF descriptors and release Matlab code. (2) We investigate the trade-off between accuracy and computational efficiency of descriptors in terms of frame sampling rate and type of Optical Flow method. (3) We investigate the trade-off between accuracy and computational efficiency for the video representation, using either a k-means or hierarchical k-means based visual vocabulary, a Random Forest based vocabulary or the Fisher kernel.

KW - Computational Efficiency, HOF, HOG, Video Classification

U2 - 10.1145/2578726.2578744

DO - 10.1145/2578726.2578744

M3 - Conference contribution

SN - 978-1-4503-2782-4

BT - Proceedings of International Conference on Multimedia Retrieval

PB - ACM

CY - New York, NY, USA

ER -

Realtime Video Classification Using Dense HOF/HOG

Abstract

Keywords / Materials (for Non-textual outputs)

Access to Document

Fingerprint

Cite this